Few patients should be readmitted for the same condition within 30 days of hospital discharge [1. Numerous approaches have been developed with success in the range of 10% to 40% reduction in readmissions including Project RED and Project BOOST. [2 [3 [4 [5. Ong and colleagues [6 used nurse calls and remote patient monitoring in patients with heart failure with no effect on 30-day OR 180-day readmissions. Stauffer [7 was successful in reducing readmissions with frequent nurse visits, but the use of nurses was prohibitively expensive [8 . A lower cost intervention that was equally or more effective was needed. GAs is 1 such possibility.
Jump to Section Methods Results Discussion
The idea of “the Grandparent Corps” was to make use of the wisdom of grandparents in home visits, until an individual from the Department of Labor pointed out that requiring an individual to be a grandparent was discriminatory against those without grandchildren. The concept of the “good grandparent” remains, creating a relationship and delivering “tough love.” GAs are nurse extenders and have had prior training either as a nurse aide, medical assistant or community health worker. Every visit is supervised in a short video conference with the patient and family (or after 2 weeks video or telephone) by an Advance Practice Registered Nurse or a Registered Nurse. The choice of the nurse supervisor is key; this person must be dedicated to the program as well as serving as a medical liaison to the physicians and Human Resources supervisor to the Grand-Aides. The Grand-Aides, most of whom are nurse aides, come from either home health agencies, nursing homes or the medical institution itself, usually from clinics.
GAs leverage supervisors with 5 GAs to 1 supervisor. Each GA can cover approximately 100 patients per year. Within 48 hours of discharge the GA makes the first home visit, with 2 more visits week 1, and then decreasing over the first month according to clinical need. The visit, video and phone frequency over the next year is determined by the clinical situation, with some patients having GA for 30 days and some longer. The procedures are detailed in detailed in Phillips [9. Patients were not participating in any other trial.
The records of all Medicare patients discharged from the University of Virginia Medical Center with a primary discharge diagnosis of heart failure between January 1, 2013 and January 1, 2015 were reviewed. Patients were excluded if they had a left ventricular assist device or were in hospice or were enrolled in Medicare Certified Home Care on discharge. Patients were offered a GA by the supervisor if they met at least 1 of the criteria for “high risk:” 1. any readmission within 30 days, 2. 2 or more admissions in the last year, 3. non-adherence with medications/ follow up appointments, 4. lives alone, and as long as they lived within 60 miles of Charlottesville, Virginia.
GAs were assigned to patients on hospital discharge. The timing for control patients began on the day of discharge for the first heart failure admission “the index admission” in that same two-year period. Data were collected on all patients at 1 month and 6 months, or earlier if the patient died.
Patient demographics and disease severity at the index admission as well as 6 month mortality are shown in Table 1. Each of these measures was collected at: 0-30 days; 31-180 days; 0-180 days (6 months). Medication adherence was assessed at 30 days using the criteria of the Adherence to Refills and Medications Scale (ARMS) test and by observation assessed by the GAs and supervisor [12 (Table 2) “Substantial adherence” was defined as levels 1 and 2. This categorization is similar to “fully adherent” used by Bansilal et al. [13
Unadjusted readmission rates for all outcomes and inverse probability of treatment weighting (IPW) adjusted odds ratios were reported for patients with an without GA. Adjustment for patient demographics and readmission risk factors (Figure 1, Figure 2) was performed via IPW using the survey package [14[15[16. Length of stay (LOS) is reported but cannot be used as a proxy for hospitalization severity because those assigned a GA may be discharged sooner than they would have been otherwise.
Standardized absolute mean differences. Covariate balance, before and after IPW weighting, shown as absolute standardized mean differences, i.e. the absolute value of the difference in means or proportions divided by a pooled standard deviation. A commonly used rule of thumb considers values < 0.1 to be well balanced.
Propensity scores (PS) were calculated via a logistic regression of Grand Aides exposure by the covariates in shown in Table 1. All covariates were binary except for age, which was fit with a restricted cubic spline. The 854 control subjects are shown in blue/purple and the 108 Grand Aides subjects are shown in red/purple. The purple overlap region represents the theoretical pair matched cohort to which both exposure groups are scaled to reflect. These matching weights are calculated like average treatment effect weights, i.e. the inverse of the probability of the exposure actually received, but with a stabilizing factor equal to the minimum of the propensity score and 1 minus the propensity score.
The program underwent expedited review and was approved by the University of Virginia Institutional Review Board.
Jump to Section Methods Results Discussion
A total of 986 Medicare patients with a primary diagnosis of heart failure were discharged between January 1, 2013 and January 1, 2015; 130 were offered the GAs program and 22 (17%) decided not to participate; the most common reason for lack of participation was not wanting to have home visits. Therefore, the remaining 83% – 108 patients – participated in the GAs program. A sensitivity analysis included the 22 decliners imagining they all had 6 month readmissions. Two potential controls were excluded for having comorbidities not present in any patients, 1 for peptic ulcer disease and 1 for AIDS, leaving 854 controls. The majority of the medication changes were done by the nurse practitioner, who had periodic check-ins with the patient's physicians. Patients continued to have their 7-10 day post-discharge clinic check-in. The overall number of clinic visits was not different between the GA and non-GA patients.
The characteristics of the patients at the time of admission in the 2 groups are shown in Table 1. Ages and genders were similar before and after weighting. Prior to weighting, there were more African American patients who received GAs than controls (43.5% vs 25.6%). Those with GAs had more severe UHC severity scores and more prior heart failure admissions. These pre-weighting imbalances are expected given that GAs were offered to those deemed at greater risk of readmission. Hence the unadjusted analyses, while statistically significant, are underestimating the benefit of GAs for heart failure related outcomes (see Methods above). LOS mean was reduced significantly in those with GAs (6.2±0.55 Standard Error of the Mean [SEM] vs 9.0±0.39, p=0.01). In the subgroups with both EF and BNP observed, more patients had levels considered abnormal at the University of Virginia: 77 vs 62% with EF < 50% and 98 vs 95% with BNP > 52 pg/mL.
Medicare patients with heart failure who had GAs had a 2.8% 30-day all-cause readmission rate compared with the control patients who received usual care in the same time period with a 15.8% 30-day all-cause readmission rate, representing an 82% reduction in readmissions (Figure 3:, Figure 4) p=0.0060. For 30-day all-cause ED visits, 2.8% of those with GAs had at least 1 visit compared with 45.1% of controls. (Figure 3:, Figure 4)
30-day all-cause readmissions and 30-day all-cause ED visits are highly significantly different comparing patients with GAs vs control patients (30-day all-cause readmissions p=0.006; ED visits p<0.0001).
Odds Ratios (95% confidence intervals) for GAs vs Control for Readmissions, ED Visits and death in the IPW Weighted Cohorts. Statistical significance is shown for the weighted analyses are denoted as * p<0.05, ** p<0.01, *** p<0.001, and **** p<0.0001. The upper bound for outcome 4a in the weighted analysis is 2.70, not shown. Abbreviations: d=day; Dept=Department; mo=month; =Patients with GAs; Non-Cardiac=Patients with cardiac readmission but not due to heart failure; Readmits=Readmissions; IPW=inverse probability of treatment weighting.
The significant reduction in all-cause readmissions persisted with a 71% reduction in 6-month all-cause readmissions (figure 5); p<0.0001. The remainder of the comparisons for readmissions and ED visits are shown in Figure 4, Figure 5:. At 6 months, each of the readmission subgroups with GAs were statistically different from control (Figure 5). The significant reduction in ED visits also persisted to 6 months with GAs 12% and control 52% (a 77% reduction) with p<0.0001.
6-month Readmissions and all-cause ED visits are highly significantly different comparing patients with GAs vs control patients.
At 6 months the death rate was for those with and without GA was not statistically different, 75/854 (8.8%) controls vs. 7/108 (6.5%) in (p=0.5327).
Several sensitivity analyses were performed. Ejection Fraction (EF) was observed on all patients, but only 46% of controls. Similarly, Brain Natriuretic Peptide (BNP) was observed in 94% of patients and 63% of controls. In the subgroup that had EF and BNP observed, both measures showed the patients to be sicker on average: mean EF 32.7 vs 41.0% (t-test p<0.0001) and median BNP 911.5 vs 709.5 pg/mL (Wilcoxon test p=0.02806). A sensitivity analysis on the subgroup yielded similar conclusions as the primary analysis: all cause readmission in 30 days 3/102 (2.9%) vs control 37/260 (14.2%) (Fisher exact p=0.0038). If we exclude from analysis the people who died between 0 and 30 days, the 30-day all-cause readmissions is then 3/102 (2.9%) for and 134/849 (15.8%) for controls, p<0.0001). At 6 months the death rate was not statistically significant, 7/108 (6.5%) vs 75/854 (8.8%) (p=0.5327). Excluding those deaths, we find for all-cause 6 month readmission 13/101 (12.9%) for GA and 320/779 (41.1%) for controls, p<0.0001). After adjusting for baseline differences via weighting, the difference in 6-month mortality rates remained not statistically significant, aOR=0.73 (0.31, 1.72) p=0.4698.
The ARMS test [12 for medication adherence was administered at 30 days (Table 2). Of the 108 patients, 82 were taking their medications “all the time.” The next criterion of “most of the time” had an additional 17 patients missing 1-2 doses per week. Therefore, 92% of patients had “substantial medication adherence” (see Methods) at 1 month.
Expenses were analyzed. The salary for a GA was $24,000 (+25% fringe = $30,000) and for a supervisor $85,000 (+25% fringe = $106,250. The supervisor cost per GA was ($106,250/5) $21,250); tablet + Internet $1,500 and transportation $12,500. Total fully allocated expense per GA per year was $65,250. The GAs USA fee for program development, training, implementation, QI, Webinars, software license, and certification / recertification) was allocated $15,000 per GA.
Jump to Section Methods Results Discussion
Patients with heart failure who had GAs had a 2.8% 30-day all-cause readmission rate compared to controls from the same time period in the same hospital of 15.8% - an 82% reduction - and had similar reductions in ED visits. While 30-day all-cause readmissions has been an important index for CMS, the finding that at 6 months, rates for all types of readmission and ED visits were statistically different with GAs, is important for patient care. The patients with GAs had higher, EF, BNP and UHC severity scores with more co-morbidities than control patients, making the difference particularly pertinent to programs that may choose to apply interventions to patients who are the most severely ill.
Numerous approaches to decreasing hospital readmissions in patients with heart failure have been developed and the results presented here appear to be among the best published data. [2[3[4 [16[17 [18[19[20 The data on telehealth are quite variable, most likely due to the type and level of support given. [21 [22 The paper from Ong [6creates question about the effectiveness of telemonitoring in patients with heart failure; even supplemented with nurse calls as there was little-to-no effect of these interventions on 30-day or 6 month readmissions, ED visits or expense.
Recently, Fonarow, et.al. [23 reported that after the Hospital Readmission Reduction Program (HRRP), (with same time period as the present study), heart failure 30-day admissions were reduced to 21.4%; 30-day mortality with HRRP was 9.2% and without HRRP was 6.6%. Our data with GAs were different from Fonarow, et.al.[23 with a 30-day readmission rate of 2.8%, and similar mortality in both the GA and control groups at 1 month and 6 months. Thus, we did not see an increase in mortality with the initiation of a readmission reduction program.
In the patients with GAs, 92% were judged to have “substantial medication adherence.” The majority of studies report that about 50% of patients take their prescribed medication as directed [24. The recently published work by Bansilal, et al. [13 demonstrates that with “fully adherent” patients (similar to “substantial adherence” in this study), patient outcomes improve. Our data support that conclusion with a relationship of improved adherence with reduced readmissions and ED visits.
The GAs expense (including all categories) was $803 per patient per year, using an estimate of 100 patients per GA per year [9. If the data from the present study are used, patients with a GA had an 82% reduction in the 30-day all-cause readmissions. However, this program had a number of positive factors, such as the same supervisor for the entire 4 years. If we even consider a decreased effectiveness by 50%, the reduction in readmissions would then be 41%. Applying 41%, as each GA had approximately 100 patients per year each likely to be readmitted, 41 readmissions would have been prevented. With a cost per readmission of $15,667 [25, the savings would have been $642,347, with an expense (using data from this study) of $80,250, the net saving per GA was $562,097 and return on Investment (ROI) was 7.0 fold.
The study has limitations. Although the time period for GA assisted patients and controls were identical and major risk factors for readmission were adjusted for, this was not a randomized trial. Patients at greater risk of readmission were offered a GA and needed to accept to be included in the study. Theoretically, the control group contains patients who would have rejected a GA had they been offered. While the inclination to refuse a GA cannot be ruled out as an independent risk factor for readmission, it would need to be an extremely strong risk factor to move the findings of this paper into statistical non-significance.[26. Second, only records from the University of Virginia were available. It could be argued that the patients were readmitted differentially to UVA as they were followed more closely. If this was the case and the controls had more ED visits and readmissions elsewhere, this would have made the results even more favorable. Finally, the GAs all had the same supervisor (contributing author on this paper). Continuity and quality of the supervisor may be an important factor in the success of a GAs program.
Where do GAs fit into the armamentarium to care for post-acute patients? The work from Bradley, et al. [27demonstrates that keeping people healthy and out of the hospital requires a multi-pronged approach. We know that different people require different types and levels of support at different times. We also know on the basis of this study that GAs are effective in providing care at relatively low cost to a broad group of patients with heart failure [28 [29. Further trials in patients with other diagnoses are underway. What needs to be determined is who needs what, when. Some patients may do well with only the support from caring family, some may do well with a “self-management” app [30, or telemedicine. We need studies not comparing intervention A with intervention B, but rather, in a population, which specific patients will get the best result from the least expensive intervention and how those needs change in an individual over time; perhaps we should call this “personalized management.” On the basis of this study, we conclude that Grand-Aides have a place in that continuum.
1. Kaiser Health News, Aiming for fewer hospital U-turns, 29 Jan 2015, pages 1-12. [Online]. Available: http://kff.org/medicare/issue-brief/aiming-for-fewer-hospital-u-turns-the-medicare-hospital-readmission-reduction-program/. [Accessed 10 October 2017].
2. Naylor MD, Brooten D, Campbell R, Maislin G, McCauley KM, Schwartz JS. Comprehensive discharge planning and home follow-up of hospitalized elders: a randomized clinical trial. JAMA, 1999; 28: 613-620.
3. Stubenrauch JM, “Project RED Reduces Hospital Readmissions.” AJN 2015; 115: 18-19.
4. James J. Medicare Hospital Readmissions Reduction Program. Health Affairs Health Policy Brief, November 12, 2013, pages 1-5.
5. Hanson, LO, Greenwald, JL, Budnitz, T, Howell, E, Halasyamani, L, Maynard, G, Vidyarthi A, Coleman EA, Williams MV. Project BOOST: Effectiveness of a multihospital effort to reduce rehospitalization. J Hosp Med2013; 8: 421-427.
6. Ong MK, Romano PS, Edgington S, Aronow HU, Auerbach AD, Black JT, De Marco T, Escarce JJ, Evangelista LS, Hanna B, niats TG10, Greenberg BH, Greenfield S, Kaplan SH, Kimchi A, Liu H, Lombardo D, Mangione CM, Sadeghi B, Sadeghi B, Sarrafzadeh M, Tong K, Fonarow GC; Better Effectiveness After Transition–Heart Failure (BEAT-HF) Research Group. Effectiveness of Remote Patient Monitoring After Discharge of Hospitalized Patients With Heart Failure. JAMA Intern Med 2016; 176:310-318.
7. Stauffer BD, Fullerton, C, Fleming, N, Ogola G, Herrin J, Stafford PM, Ballard DJ. Effectiveness and cost of a transitional care program for heart failure: a prospective study with concurrent controls. Arch Intern Med2011; 171: 1238-1243.
8. Katz, M. “Interventions to decrease hospital readmission rates: Who saves? Who pays? Arch Intern Med2011; 171: 1230-1231.
9. Garson, A. Leveraging the workforce for more effective and less expensive care. in America's Healthcare Transformation: Strategies and Innovations. In: Phillips, RA, ed. New Brunswick: Rutgers University Press, 2016: 178-198.
10. Averill, RF, Goldfield, N, Steinbeck, B, Grant, T, Muldoon, J, Brough, AJ, Gay, J. Development of the All Patient Refined DRGs. 3M HIS Research Report 8-97, 1997, pages 1-22. [Online]. Available: http://solutions.3m.com/3MContentRetrievalAPI/BlobServlet?lmd=1225920498000&assetId=1180606514429&assetType=MMM_Image&blobAttribute=ImageFile. [Accessed 10 October 2017]
11. American College of Surgeons, “Charlson/Deyo Score,” National Cancer Data Base - Data Dictionary PUF 2015, 1 Jan 2014, pages 1-2. [Online]. Available: http://ncdbpuf.facs.org/content/charlsondeyo-comorbidity-index. [Accessed 10 October 2017].
12. Kripalani S, Risser J, tti ME, Jacobson TA. Development and evaluation of the Adherence to Refills and Medications Scale (ARMS) among low-literacy patients with chronic disease. Value Health 2009; 12 :118-123.
13. Bansilal S, Castellano JM, rrido E, Wei HG, Freeman A, Spettell C, rcia-Alonso F, Lizano I, Arnold RJ, Rajda J, Steinberg G, Fuster V6. Assessing the impact of medication adherence on long-term cardiovascular outcomes. J Am Coll Cardiol 2016; 68: 789-801.
14. Li L, Greene T. A weighting analogue to pair matching in propensity score analysis. Int J Biostat 2013; 9: 215-34.
15. Lumley, T.: Package ‘survey,’ Survey Packages, 2017, pages 14-30. [Online] Available: https://cran.r-project.org/web/packages/survey/survey.pdf. [Accessed 10 October 2017].
16. R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, 2015, pages 6-13.. [Online] Available: http://www.R-project.org/ [Accessed 10 October 2017].
17. Hostetter M, Klein, S. “Avoiding Preventable Hospital Readmissions by Filling in Gaps in Care: The Community-Based Care Transitions Program,” The Commonwealth Fund Quality Matters Archive, August/September 2012, pages 1-7. [Online]. Available: http://www.commonwealthfund.org/publications/newsletters/quality-matters/2012/august-sept/in-focus. [Accessed 10 October 2017].
18. Voss R, rdner R, Baier R, Butterfield K, Lehrman S, Gravenstein S. The care transitions intervention: translating from efficacy to effectiveness. Arch Intern Med 2011; 171: 1232-7.
19. Coleman EA, Parry C, Chalmers S, Min SJ. The care transitions intervention: results of a randomized controlled trial. Arch Intern Med 2006; 166: 1822-1828.
20. Kangovi, S, Grande, D, Chau T-S. From Rhetoric to Reality — Community Health Workers in Post-Reform U.S. Health Care. N Engl J Med 2015; 372:2277-2279.
21. Desai, AS. Home Monitoring Heart Failure Care Does Not Improve Patient Outcomes: Looking Beyond Telephone-Based Disease Management. Circulation 2012;125: 828-836.
22. Maeng, DD, Starr, AE, Tomcavage, JF, Sciandra, J, Salek, D, Griffith, D. Can Telemonitoring Reduce Hospitalization and Cost of Care? A Health Plan's Experience in Managing Patients with Heart Failure. Population Health Management 2014; 17: 340-344.
23. Fonarow, GC, Konstam, MA, Yancy, CW, The hospital readmission program is associated with fewer readmissions, more deaths: Time to reconsider. J Am Coll Cardiol 2017; 70: 1931-33.
24. Centers for Disease Control. “ Overcoming Barriers to Medication Adherence for Chronic Diseases.” Public Health Grand Rounds 2017 pages 1-2. [Online]. Available: https://www.cdc.gov/cdcgrandrounds/pdf/archives/2017/GR_02-21-2017.pdf. [Accessed 10 October 2017]
25. Fitch K, “The high cost of heart failure for the Medicare population: An actuarial cost analysis.” Milliman Report, 2017, pages 1-29 [Online]. Available http://us.milliman.com/insight/2017/The-cost-burden-of-worsening-heart-failure-in-the-Medicare-fee-for-service-population-An-actuarial-analysis/ [Accessed 10 October 2017]
26. Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics 1998: 54: 948-963.
27. Bradley EH, Curry L, Horwitz LI, Sipsma H, Wang Y, Walsh MN, Goldmann D, White N, Piña IL, Krumholz HM. Hospital strategies associated with 30-day readmission rates for patients with heart failure. Circ Cardiovasc Qual Outcomes 2013; 6: 444-50.
28. Storrow AB, Jenkins CA, Self WH, Alexander PT, Barrett TW, Han JH, McNaughton CD, Heavrin BS, Gheorghiade M, Collins SP. The burden of acute heart failure on U.S. emergency departments. JACC Heart Fail 2014; 2: 269-77.
29. Voigt J, Sasha John M, Taylor A, Krucoff M, Reynolds MR, Michael Gibson C. A reevaluation of the costs of heart failure and its implications for allocation of health resources in the United States. Clin Cardiol 2014; 37:312-21.
30. Wike, H. “Give patients tablets, Reduce Readmissions,” Health IT Outcomes 2013, page 1. [Online]. Available: https://www.healthitoutcomes.com/doc/give-patients-tablets-reduce-readmissions-0001.[Accessed 10 October 2017].