Table 1.

Warfarin indications, rate of use and prescribing patterns

Study	Study objective, (intervention/exposure and outcomes)	Study design, data source	Study population, study setting, time period	Results	Quality assessment, funding source
Part A: Warfarin exposure
Abdel-Latif et al. (2005) [7]	To determine predictors of OAC therapy for AF in LTC	Design: case control study Data source: medical chart, pharmacy record, and MDS	Population: 117 residents with chronic or paroxysmal AF identified from 934 total residents Setting: 6 LTC facilities with > 100 beds in Cleveland metropolitan area (US) Time period: not specified	Among 117 residents (12.5% of 934) with AF, OAC was prescribed for 46%; aspirin or clopidogrel: 40%; no antithrombotic treatment: 21%. Logistic regression produced 2 independent predictors of OAC prescription: (1) Prior stroke was the primary determinant of receiving OAC (OR = 0.02; 95% CI = 0.09-0.47) sic, and (2) history of GI bleeding was a predictor for not receiving OAC (OR, 5.6; 95% CI = 1.1- 29.4). Classification and regression tree analysis found residents with prior stroke or GI bleeding and no history of coronary artery disease and who were non-Caucasian were less likely to be prescribed OAC. Those without stroke were less likely to be prescribed OAC if they were younger, had dementia or lower functional status	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? No 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes; date of study not specified; cross-sectional 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Yes 11) Analytic methods appropriate? Partial; MDS cognition and functional scale scoring methods not referenced; multivariate findings not fully reported Funding: not specified
Christian et al. (2003) [17]	To evaluate the extent to which people of color (e.g. non-white or Hispanic) in US nursing homes were less likely to receive pharmacologic treatment of recurrent stroke	Design: retrospective cross-sectional study Data source: the SAGE database (links inpatient Medicare claims, drug data, and MDS data)	Population: 19,051 nursing home residents with recent hospitalization for ischemic stroke Setting: Kansas, Maine, Mississippi, Ohio, New York, and South Dakota (US) Time period: 1992-1996	Variability in use of any treatment for secondary stroke prevention (warfarin or antiplatelet agent) was observed by race/ethnicity: 58% of American Indians received therapy, 54% of non-Hispanic whites, 49% of non-Hispanic blacks, 46% of Hispanics, and only 39% of Asian/Pacific Islanders. The use of warfarin among residents with conditions warranting anticoagulant therapy was 40% among non-Hispanic whites, 36% among American Indians, 32% among non-Hispanic blacks, 26% of Asian/Pacific Islanders, and 25% among Hispanics. After controlling for confounding, Asian/Pacific Islanders (prevalence difference = -5.2, CI = -18.1 to 7.8), non-Hispanic black residents (prevalence difference = -7.6, CI = -11.2 to -3.9), and Hispanics (prevalence difference = -7.6, CI = -17.6 to 2.2) received warfarin less often than non-Hispanic whites	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Partial; likely lacked statistical power for Asian/Pacific Islanders, Hispanics, and American Indians 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? No; uses data from admission only so secondary stroke prevention was more likely ordered at hospital discharge, not in LTC facility 9) Completeness of follow-up? No 10) Analysis controls for confounding? Yes 11) Analytic methods appropriate? Partial; prevalence differences reported rather than association strength (i.e. no OR) Funding: National Institute on Aging, AHRQ
Gurwitz et al. (2007) [25]	To examine the preventability of actual and potential warfarin-related adverse events in the nursing home setting	Design: retrospective cohort study Data source: nursing home records	Population: all 490 residents of 25 nursing homes receiving warfarin therapy Setting: 25 nursing homes (bed size range, 90-360) in Connecticut (US) Time period: 12- month observation period (Apr 2003 - Mar 2004)	The most common indications for warfarin therapy included stroke prevention in AF (58%), treatment/prevention of DVT or PE (26%), and stroke prevention without AF (12%)	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Partial; classification of warfarin use/non-use within a given resident-month not explicated 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? No; Did not analyze prognostic factors associated with adverse warfarin-related events other than warfarin exposure 11) Analytic methods appropriate? Partial; association of resident characteristics or INR values with risk of adverse warfarin-related event not assessed Funding: AHRQ
Gurwitz et al. (1997) [8]	To determine the prevalence of AF in the institutionalized elderly population and the proportion receiving warfarin; to identify clinical and functional characteristics of institutionalized elderly persons with AF that are associated with the use of warfarin; access quality of warfarin prescribing and monitoring	Design: retrospective cohort study Data source: medical record review of residents with ≥ 2 weeks of warfarin therapy during the 12-month period preceding the date of medical record abstraction	Population: 6437 residents of LTC facilities Setting: 30 LTC facilities (6437 total number of beds) located in New England, Quebec, and Ontario (US and Canada) Time period: Jul 1993-Aug 1995	An electrocardiogram indicating AF was present in the records of 7.5% of 5500 LTC residents; 32% of such patients were being treated with warfarin. In multivariate analysis, only a history of stroke (OR = 1.87; 95% CI = 1.20-2.91) was found to be positively associated with the use of warfarin in this setting. Patients with a diagnosis of dementia (OR = 0.59; 95% CI = 0.38-0.90) and those aged _85 years (OR = 0.46; 95% CI = 0.22-0.94) were less likely to receive warfarin therapy. Warfarin was commonly prescribed to patients with a history of bleeding (28.5%), substantial co-morbidity (30.8% major) and functional impairment (25.4% severe), a history of falls (28.5%), or concomitant potentiating drug therapy (17.7%)	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes (cross-sectional) 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Yes 11) Analytic methods appropriate? Yes Funding: DuPont Pharma, the National Institute on Aging, Medical Research Council of the Province of Quebec
Hughes et al. (2004) [18]	To identify factors relating to initiation and discontinuation of secondary stroke prevention agents (warfarin and antiplatelets) among stroke survivors in nursing homes.	Design: retrospective cohort study Data source: MDS for patient characteristics matched to the OSCAR system for facility characteristics	Population: 16,579 stroke survivors; 9547 were not receiving any secondary stroke prevention treatment at admission; 6244 were receiving therapy Setting: nursing homes in 6 states (US) Time period: 1992-1996: each resident was followed ≥ 6 months	In all, 12% initiated drug therapy (warfarin or antiplatelet); 30.3% discontinued. Conditions known to increase the risk of recurrent stroke (e.g. AF) were predictive of initiation. Factors inversely related to initiation of therapy included advanced age, severe cognitive impairment, and being dependent in ADLs. Co-morbid conditions were inversely related to discontinuation of treatment, whereas advanced age and severe cognitive impairment increased likelihood of discontinuation. Black residents (OR = 0.62; 95% CI = 0.49-0.78) were less likely than non- Hispanic white residents to initiate therapy. Asian/Pacific Islanders (OR = 0.44; 95% CI = 0.23-0.83) were less likely than non-Hispanic white residents to discontinue therapy	Quality assessment: for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Partial; range of follow-up for observing initiation/discontinuation events was 6-13 months 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Partial; length of follow-up not treated as a covariate in adjusted logistic regression models 11) Analytic methods appropriate? Yes Funding: National Institute on Aging
Lackner et al. (1995) [9]	To assess warfarin use and monitoring in nursing home patients with NVAF, according to American College of Chest Physicians Consensus Conference guidelines	Design: retrospective, cross-sectional study Data source: medical record review and attending physician response to written communication from the nursing home's medical director and consultant pharmacist	Population: 902 patients aged ≥ 60 years, from whom 69 with a diagnosis of NVAF and 16 with VAF (control group) were identified Setting: 5 nursing homes in Minneapolis-St Paul, Minn (US) Time period: 1-month period (June 1993)	NVAF was documented in 7.6% and VAF in 1.8% of the patients. Only 17% of patients with NVAF were receiving warfarin, compared to 31% of patients with VAF. 58% of patients with NVAF and without a conventional contraindication to warfarin had ≥ 1 risk factor for thromboembolism in addition to AF and advanced age, yet only 20% used warfarin	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? No; low power 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes (cross-sectional) 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Partial; evaluated univariate association of stroke risk factors and warfarin contraindications with warfarin use 11) Analytic methods appropriate? Partial; CIs not included in any findings Funding: Dupont Pharmaceuticals
Lapane et al. (2006) [19]	To evaluate the impact of the implementation of the Medicare PPS on pharmacologic secondary ischemic stroke prevention (standing orders for antiplatelets or warfarin) in nursing homes	Design: retrospective cohort study Data source: the SAGE database (including MDS data and all drugs taken 7 days preceding MDS assessment)	Population: residents who were hospitalized with an ischemic stroke within 6 months (1997, n = 5008; 2000, n = 5243) of living in nursing facilities Setting: nursing facilities in Kansas, Maine, Mississippi, or Ohio (1997: n = 1226; 2000: n = 1092) (US) Time period: Pre- PPS period = 1997; post-PPS period = 2000	The unadjusted proportion of use of pharmacologic agents for the secondary prevention of stroke was similar for warfarin in both time periods (1997: 22.9%; 2000: 22.4%) and increased for antiplatelets (1997:40.8%; 2000: 47.7%), as a result of the introduction of clopidogrel. Among residents with conditions indicating the use of warfarin, after adjusting for resident and facility characteristics, the likelihood of use of antiplatelets increased in the post-PPS era (adjusted OR = 1.26; 95% CI = 1.15-1.38); the likelihood of use of the use of warfarin did not change (adjusted OR = 0.99; 95% CI = 0.86-1.14)	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Partial; design does control for effects of history other than implementation of PPS between pre- and post-PPS period (e.g. issuance of prescribing guidelines) 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Partial; evaluated only 1 pre- and post-PPS year; cross-sectional 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Yes 11) Analytic methods appropriate? Yes Funding: supported by a National Primary Care Career Scientist Award from the Research and Development Office, Northern Ireland
Lau et al. (2004) [10]	To identify patterns and predictors of antithrombotic use and to evaluate the appropriateness of antithrombotic therapy for stroke prophylaxis in institutionalized elderly patients with AF	Design: retrospective cohort study Data source: Administrative databases and medical records	Population: 265 LTC residents, aged ≥ 65 and older, with AF Setting: 17 LTC institutions in Edmonton, Alberta (Canada) Time period: Nov 2001 - Feb 2002	Warfarin was prescribed for 49% of patients, aspirin for 22%, both for 8%, and neither for 20%. Nearly all patients (97%) were considered to be at high risk for stroke, with age being the predominant risk factor (88% ≥ 75 years), whereas about half (54%) were considered to be at low risk for bleeding. Multivariate analyses found no associations between individual risk factors for bleeding and anticoagulation treatment, with the exception of recent surgery (OR = 0.59; 95% CI = 0.37-0.94). Overall, 54.8% of patients received appropriate antithrombotic therapy congruent with stroke and bleeding. Of patients who were optimal candidates for anticoagulation, 60% received appropriate therapy (warfarin with or without aspirin)	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Partial; unlike stroke risk, categorization by bleeding risk not based on validated algorithm or consensus guideline 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes; cross-sectional 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Yes 11) Analytic methods appropriate? Yes Funding: not specified
McCormick et al. (2001) [11]	To assess: (1) the prevalence of AF and the percentage of AF patients who receive therapy with warfarin or aspirin, (2) the relationship between the presence of known risk factors for stroke and bleeding among persons with AF and their receipt of warfarin, and (3) the quality of warfarin prescribing and monitoring in nursing home residents with AF	Design: retrospective cohort study Data source: Medical record review	Population: 2587 LTC residents Setting: 21 LTC facilities in Connecticut (US) Time period: 1997-1998	AF was present in 17% of LTC residents, risk factors for stroke in 93% of AF residents, and for bleeding in 80% of AF residents. Overall, 42% of AF patients were receiving warfarin. However, of 83 ideal candidates, only 53% were receiving this therapy. The odds of receiving warfarin in the study sample decreased with increasing number of risk factors for bleeding (adjusted OR for > 1 bleeding risk factor compared to none: 0.51; CI, 0.29-0.94) and increased (non-significant trend) with increasing number of stroke risk factors	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? No; stroke risk and bleeding risk classification not adequately described (i.e. no reference to validated algorithm or consensus guideline) 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes; cross-sectional 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Partial; a small list of potential confounders was included in the logistic regression model 11) Analytic methods appropriate? Yes Funding: Health Care Financing Administration, Department of Health and Human Services
Quilliam et al. (2001) [12]	To explore characteristics of nursing home residents who are stroke survivors and factors associated with secondary prevention of stroke in nursing homes	Design: Retrospective cross-sectional study Data source: MDS	Population: 53,829 (20.4%) residents aged > 65 years with a diagnosis of stroke (stroke type unknown) Setting: all nursing home residents in 5 states (US) Time period: 1992-1995	67% of stroke survivors and > 50% of those hospitalized with stroke over the previous 6 months were not receiving drug therapy for stroke prevention. Among those treated, most received aspirin alone (16%) or warfarin alone (10%). Independent predictors of drug treatment included co-morbid conditions (e.g. hypertension, AF, depression, Alzheimer's disease, dementia, history of GI bleeding, and peptic ulcer disease). Those aged ≥ 85 years were less likely to be treated than those aged 65-74 years (OR = 0.86; 95% CI = 0.82-0.91); black residents were less likely to be treated than whites (OR = 0.80; 95% CI = 0.75-0.85); and those with severe cognitive (OR = 0.63; 95% CI = 0.60-0.67) or physical impairment (OR = 0.69; 95% CI = 0.64-0.75) were also less likely to receive drug treatment	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Partial; only a limited set of bleeding risk factors were considered in the logistic regression model 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes; cross-sectional 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Yes 11) Analytic methods appropriate? Yes Funding: National Institute on Aging, AHRQ
Sloane et al. (2004) [13]	To determine the prevalence and predictors of non-prescribing of selected medications for 4 common geriatric conditions (including aspirin or anticoagulants for persons with a history of stroke) whose value in decreasing morbidity has been established in clinical trials	Design: Cross-sectional study Data source: patient characteristics and diagnoses were based on medical record reviews and in-person patient assessments; data on facility characteristics were obtained by interviewing facility administrators	Population: 2014 residents aged ≥ 65 years Setting: a stratified random sample of 193 residential care/assisted living facilities in Florida, Maryland, New Jersey, and North Carolina (US) Time period: Oct 1997 - Nov 1998	Of 435 patients with prior stroke (stroke type not specified) 14.4% had a contraindication for aspirin use and 0% had a contraindication for warfarin use. 37.5% were not receiving an anticoagulant or antiplatelet agent. Neither bivariate nor multivariate analysis showed an association between non-prescribing and resident characteristics. Some facility characteristics were associated with non-prescribing in bivariate analysis (traditional vs small facility [OR = 0.55; P < 0.05], new model vs small facility [OR = 0.47; P < 0.01], presence of an RN/LPN [OR = 0.58; P < 0.05]). However, in the multivariate analysis no facility characteristics were significantly associated with Non-prescribing	Quality assessment for observational studies: 1) Unbiased selection of the cohort? Yes 2) Selection minimizes baseline differences in prognostic factors? Yes 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? No; contraindication to warfarin use not evaluated; contraindication to aspirin use limited to peptic ulcer disease 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? Yes 8) Adequate follow-up period? Yes; cross-sectional 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? Partial; many covariates in multiple drug therapy study have little relevance to warfarin or antiplatelet use 11) Analytic methods appropriate? Yes Funding: National Institute on Aging
Part B: Medication management interventions
Crotty et al. (2004) [20]	To assess whether pharmacist outreach visits would improve the implementation of evidence-based clinical practice in the area of falls reduction and stroke prevention in a residential care setting	Design: randomized control trial Data source: pre- and post-intervention case note audits	Population: 452 residential care staff was surveyed; 121 physicians were involved, with 61 receiving outreach visits. Pre- and post-intervention data were available for 715 LTC residents Setting: 10 nursing homes and 10 hostels (low-level facilities) in South Australia Time period: 7-month follow-up period (dates not specified)	No statistically significant difference between groups for numbers of patients at risk of stroke on aspirin at follow-up. Percent of residents with AF recorded on warfarin was similar between groups: 22.6% (pre) and 17.1% (post) in the control group, and 8.6% (pre) and 16.7% (post) in the intervention group (RR = 0.92; 95% CI = 0.23-3.95)	Quality assessment for observational studies: 1) Baseline comparability? Yes 2) Valid AD/cognitive outcomes assessment? Yes 3) Subjects/providers blind? Cannot be determined 4) Outcome assessors blind? Yes 5) Incomplete data adequately addressed? Yes 6) Differential dropout rate < 10%? Yes 7) Overall dropout rate < 30%? Yes; was 22.5% but as high as 37% in some cluster facilities 8) Conflict of interest reported and insignificant? Yes 9) Randomization adequate? Yes 10) Allocation concealment adequate? Yes Funding: National Health & Medical Research Council Evidence Based Clinical Practice Research Program
Horning et al. (2007) [21]	To evaluate clinical practice guideline adherence (including antiplatelet and anticoagulation therapy for secondary stroke in prevention) in patients LTC facilities who received pharmacist-directed DSM compared with patients in other LTC facilities who received traditional DRR	Design: retrospective cohort study Data source: chart review	Population: for the secondary stroke prevention subgroup, 18 stroke patients who received DSM services and 86 stroke patients who received DRR services Setting: DSM services (intervention) in 2 LTC facilities and DRR services (control) in 4 LTC facilities (US) Time period: Nov 2005	For patients with prior stroke, more DSM vs DDR patients received aspirin, clopidogrel or warfarin or were recognized with a contraindication (unadjusted, 88.9% vs 69.8%; P = 0.096; adjusted OR = 5.380; 95% CI = 0.975- 29.684)	Quality assessment for observational studies: 1) Unbiased selection of the cohort? No; control group was determined retrospectively 2) Selection minimizes baseline differences in prognostic factors? Partial; control group was determined by authors to be representative mix of local usual pharmacist consultant services 3) Sample size calculated/5% difference? No; intervention group may have lacked power due to low n = 107 residents 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Partial; description of DSM intervention for each of seven diseases evaluated was limited 6) Validated method for ascertaining clinical outcomes? Partial; description of guideline adherence scoring limited; only cite consensual guidelines 7) Outcome assessment blind to exposure? No 8) Adequate follow-up period? Yes 9) Completeness of follow-up? Yes 10) Analysis controls for confounding? No; did not adjust for institutional characteristics in logistic regression models 11) Analytic methods appropriate? No; limited covariate set unchanged among seven disease logistic regression models; stroke and bleeding risk not adequately modeled; no multiplicity adjustment Funding: No outside funding.
Papaioannou et al. (2010) [24]	To evaluate the MEDeINR system (an electronic decision support system based on a validated algorithm for warfarin dosing) by examining the impact on INR control, testing frequency, and experiences of staff in using the system	Design: retrospective/prospective cohort study (pre-post implementation design) Data source: pre-implementation: retrospective chart audit; post-implementation: central computer database	Population: 128 residents (without prosthetic valve) who were taking warfarin Setting: 6 LTC homes in Ontario (Canada) Time period: 6 months, 3 months prior to MEDeINR implementation and 3 months post-implementation (dates not specified)	128 (10%) of all residents (excluding those with a prosthetic valve) were taking warfarin in 6 LTC homes. The primary indications for taking warfarin were: AF (74%), DVT (20%), and PE (6%)	Quality assessment for observational studies: 1) Unbiased selection of the cohort? No; potential survivor bias since residents who discontinued warfarin prior to intervention due to poor INR control would not have been eligible for study 2) Selection minimizes baseline differences in prognostic factors? Partial; pre- and post-intervention without control does not adjust for biases such as history and maturation 3) Sample size calculated/5% difference? Yes 4) Adequate description of the cohort? Yes 5) Validated method for ascertaining exposure? Yes 6) Validated method for ascertaining clinical outcomes? Yes 7) Outcome assessment blind to exposure? No 8) Adequate follow-up period? Partial; some residents had < 3 months of follow-up 9) Completeness of follow-up? No; differential follow-up in pre- and post- periods not evaluated 10) Analysis controls for confounding? Partial; no covariates modeled since subjects served as own controls; assumes no time-varying relevant covariates 11) Analytic methods appropriate? Partial; sensitivity analysis to test survivor bias not performed Funding: Canadian Institute of Health Research

ADL, activities of daily living; AF, atrial fibrillation; AHRQ, Agency for Healthcare Research and Quality; CI, confidence interval; DSM, disease state management; DRR, drug regimen review; DVT, deep vein thrombosis; GI, gastrointestinal; INR, international normalized ratio; LTC, long-term care; LPN, licensed practical nurse; MDS, Minimum Data Set; NVAF, nonvalvular atrial fibrillation; OAC, oral anticoagulation; OR, odds ratio; OSCAR, Online Survey Certification and Automated Record; PE, pulmonary embolism; PPS, prospective payment system; RN, registered nurse; RR, relative risk; SAGE, Systematic Assessment of Geriatric drug use via Epidemiology; VAF, valvular atrial fibrillation