A hospital-based transitional care program for patients with heart failure or pulmonary disease failed to reduce 30- or 90-day readmissions or emergency department visits.
Medicare penalizes hospitals with 30-day readmissions above their expected rates. Hospitals have responded by implementing transitional care interventions; however, there is limited evidence to inform the development of a successful intervention.
Parallel-group, stratified, randomized controlled trial.
A total of 512 patients hospitalized at 2 community hospitals, with congestive heart failure (CHF) or chronic obstructive pulmonary disease (COPD), were randomly assigned to the intervention (n = 253) or usual care (n = 259). The intervention encompassed a 90-day hospital-based transitional care program. The primary end points were 30- and 90-day all-cause readmissions. Secondary measures included all-cause emergency department (ED) visits and mortality.
On average, study participants were 67 years of age, 57% female, and 70% insured by Medicare. There was no statistical difference between treatment groups on 30-day readmission incidence rates (difference, 0.040; 95% CI, —0.047 to 0.127; P = .36), or 90-day readmission incidence rates (difference of 0.035; 95% CI –0.122 to 0.192; P = .66). Groups also did not differ in ED visit incidence rates at 30 or 90 days. The mortality rate among patients with CHF showed no difference between groups (risk ratio = 0.90; 95% CI, 0.40-2.05). However, for COPD, mortality at 90 days was lower in the intervention group than in the usual care group (risk ratio = 0.28; 95% CI, 0.10-0.83).
Stand-alone community hospitals may be unable to prevent readmissions despite the use of comprehensive, evidence-based intervention components that are within their control. Better collaboration between hospitals and community-based providers is needed to ensure continuity of care for discharged patients.
Trial Registration: ClinicalTrials.gov, Identifier: NCT01855022
Am J Manag Care. 2014;20(10):783-792
CMS is implementing an array of policies intended to reduce readmissions because they are both costly and thought to be mostly avoidable.1 The CMS approach includes both penalties for hospitals if 30-day readmission rates for specified conditions exceed expected risk-adjusted rates, and support for demonstration projects aimed to enhance the patient’s transition from hospital to community.2 The commercial sector is also signaling a move in a similar direction. With the addition of all-cause readmission rates to the Healthcare Effectiveness Data and Information Set (HEDIS), health plans are positioned to create new incentives for hospitals to reduce readmissions.3
While hospitals have responded by implementing interventions to reduce 30-day readmissions, existing empirical evidence offers limited guidance on how to develop a successful program. Out of 43 studies in a recent systematic review of interventions to reduce readmissions, only 16 were deemed effective, and no single intervention component was consistently associated with effectiveness.4 Moreover, many of these interventions were developed and tested in academic settings or integrated delivery systems, leaving open the question of how even effective interventions would fare in stand-alone community-based hospitals. The lack of proven successful and generalizable models argues for continued innovation and experimentation with interventions designed to reduce readmissions.
In this study, we examined whether a comprehensive hospital-based transitional care intervention reduces readmissions for participants with congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD). We selected these conditions because they represent 2 of the top 3 readmitting diagnoses among Medicare beneficiaries.1 In addition, hospitals are currently penalized by CMS for 30-day readmissions among CHF patients, and COPD will be included as a measure in 2015. We hypothesized that rates of readmission would be lower for participants receiving the intervention, as would the secondary outcomes of emergency department (ED) visits and mortality rates.
METHODSTrial Design and Setting
We conducted a parallel-group, stratified, randomized controlled trial at 2 independent, nonprofit hospitals: a 124-bed acute care facility located in Grants Pass, Oregon, and a 387-bed full-service tertiary facility in Medford, Oregon. The 2 hospitals serve 9 counties in southern Oregon and northern California (total population 574,214 in 2012) and had combined annual acute discharges of 21,430 in 2012. The Southern Oregon Institutional Review Board approved the study protocol and provided oversight for the duration of the trial.
Study participants were enrolled between June 2010 and November 2012, with follow-up activities provided until the end of January 2013. During the enrollment period, the hospital census report was reviewed daily at each hospital to identify hospitalized patients with a primary diagnosis of CHF or COPD using codes from the International Classification of Diseases, 9th Revision, Clinical Modification. Patients were eligible for inclusion if they: were at least 18 years of age; resided in any of the 9 counties served by the hospitals; were community-dwelling; had sufficient cognitive ability to communicate and reason (determined by study nurse); had access to a telephone; could communicate in English; and were not currently participating in another intervention aimed at reducing readmissions.
Prior to study commencement, the principal investigator (AL) generated a randomization sequence to allocate participants to treatment arms using random permuted blocks.5 There were 4 strata (2 hospitals and 2 disease conditions) and 5 permuted blocks allocated in equal proportions, with a minimum size of 2 and maximum size of 10. The treatments were allocated in a 1:1 ratio, with 18 extra allocations provided to maintain the integrity of the final block in each stratum. The allocation sequence was concealed via sequentially numbered, opaque sealed envelopes. After participants signed the consent form and provided baseline information, the envelope was opened by study staff in their presence, simultaneously revealing the treatment allocation to both participant and study staff.
The intervention included a comprehensive set of components commonly found in the transitional care model.6 Following the taxonomy proposed by Hansen et al,4 these components included: 1) pre-discharge components: patient education, discharge planning, medication reconciliation, follow-up appointment scheduled; 2) post discharge components: timely follow-up, follow-up telephone call, availability of patient hotline; and 3) bridging components: transition/health coach, patient-centered discharge instructions.
To enhance the innovative nature of the intervention, 2 post discharge components were added—motivational interviewing—based health coaching (MI) and symptom monitoring using interactive voice response (IVR). MI is a standardized, evidence-based health coaching approach described as a “collaborative, goal-oriented style of communication with particular attention to the language of change.”7 MI and IVR were selected because the literature supports their efficacy for: coaching in short duration interventions8; promoting activation for self-management9,10; addressing challenging behavior changes such as treatment compliance and smoking cessation11,12; and reducing hospitalizations for both CHF and COPD.4,13,14 (A complete description of intervention components is provided in .)
The MI component included 1 initial session from a study nurse during the index hospital stay to prepare participants to transition back to the community setting. A follow-up session was scheduled within 2 days of discharge, with additional sessions scheduled during the subsequent 90-day period based on the participant’s patient activation level, health literacy, severity of health condition, and preference. All sessions were delivered by 1 of 2 experienced nurses trained in administering transitional care and MI. The nurses were regularly monitored by an MI expert (SWB), using the Motivational Interviewing Treatment Integrity (MITI) tool, to ensure proficiency levels in MI associated with positive clinical outcomes.15 From multiple random checks, the average global clinician ratings for the 2 nurse-coaches were 3.9 and 4.2 out of 5 (with 3.5 or above indicating proficiency); their average percentage of using MI-adherent strategies were 91% and 99% (with 90% or above indicating proficiency).
The IVR component included daily symptom monitoring using an IVR system (Pharos Innovations, Northfield, Illinois) up to 30 days post discharge, with the 2 study nurses responding to symptom alerts within 24 hours.
Although the Transitional Care Model sometimes includes home visits, we did not include this in the intervention due to funding constraints and the lack of evidence that it is a compelling component. We also could not ensure provider continuity of care, given that the funding was restricted to hospital staff, and community-based providers were not obligated to participate in this hospital-based intervention.
Participants randomized to the control condition received usual care from the hospital, with transitional care typically consisting of brief patient education and discharge planning delivered in the traditional medical model.16
Data and Outcomes
The primary outcomes, 30- and 90-day all-cause readmissions, were created from the hospitals’ electronic medical record (EMR) system data. Secondary outcomes included: 30- and 90-day all-cause ED visits, also retrieved from the hospitals’ EMRs; and mortality rates at 90 days, obtained by linking patient Social Security numbers to the Social Security Death Index and searching obituaries in the regional newspapers. As suggested during the peer review process, we also created a composite rate variable that captured all unplanned returns to the hospital, composed of both readmissions and ED visits.
Self-reported sociodemographic characteristics, comorbidities, severity level of the primary condition (based on New York Heart Association [NYHA] functional classification for CHF and Global Initiative for Chronic Obstructive Lung Disease [GOLD] classification for COPD),17 and contact information of the participant’s community-based primary care provider (PCP) were collected during recruitment. Participants also completed the Patient Activation Measure (PAM) during recruitment (baseline) and by mail at 30- and 90-day time points. The PAM assesses knowledge, skills, beliefs, and behaviors needed to enable patients to manage their health and care.18,19 To assess the level of patient engagement with the novel components of our intervention (MI and IVR), we collected the following process measures: the number of MI sessions completed, IVR usage, and the relationship between IVR alert triggers and subsequent acute care utilization.
Sample size calculations were estimated for over-dispersed Poisson observations,20 based on the hospitals’ prior 90-day readmission incidence rate of 0.51 per person. We calculated sample size separately for each condition to enable sufficient power to conduct subgroup analyses, resulting in sample sizes of 119 participants per group needed to detect a 0.23 incidence rate reduction of readmissions during a 90-day time period (2-tailed, α = 0.05, power = 0.80). The projected intervention effect size was based on successful randomized trials with similar emphases, follow-up periods, and sample sizes.21-24
All analyses were performed on the whole sample and separately by condition. Analyses were conducted from an intention-to-treat perspective. Baseline comparability of the intervention and usual care groups was evaluated using χ2 tests for categorical variables and t tests for continuous variables. We assessed the impact of the intervention on readmission incidence rates, ED visit incidence rates, and the composite rate using negative binomial regression. This model was chosen based on goodness of fit statistics that favored negative binomial over other count models.25 A count model is the most appropriate statistical approach given that some patients may experience multiple readmissions, and the data are expected to be highly skewed. Methods that recast the outcome as a binary variable, such as time-to-event analysis, introduce measurement bias by masking potential differences between study groups in their distribution of readmissions.
Differences in mortality rates between treatment and control groups were estimated using logistic regression and calculated as both risk differences and risk ratios, using the adjustment method proposed by Norton et al.26
The analysis plan originally called for adjustment to outcome models if significant imbalances on baseline characteristics were observed, or if important prognostic variables were identified. While no such concerns were substantiated, we nonetheless present all outcomes as both unadjusted and adjusted for all baseline covariates. For all analyses, P values were based on 2-tailed tests with values less than .05 considered statistically significant. Regression-based analyses estimated 95% confidence intervals using robust standard errors. All statistical analyses were conducted using Stata version 12.1 (Statacorp LP, College Station, Texas).
Of 566 patients with CHF assessed for eligibility, 277 met inclusion criteria, of whom 93% (n = 257) agreed to participate and provided baseline data (). Of these, 128 patients were assigned to usual care and 129 patients were assigned to the intervention. Among 404 screened COPD patients, 276 met inclusion criteria, of whom 255 (92%) agreed to participate and were randomized to either usual care (n = 131) or intervention (n = 124).
Intervention and usual care groups were comparable on all baseline demographic and health characteristics, including illness severity of the primary condition, comorbidities, and prior 12-month hospital utilization. This was true for the pooled study population () as well as for the CHF and COPD groups, independently ( and , available at www.ajmc.com). In general, participants were predominately female, married (or living with a caregiver), older than 65 years, insured by Medicare, and ill with several comorbidities. Participants had also utilized substantial acute hospital services in the prior year.
Process Measures: MI and IVR
Intervention group participants had an average of 6 (SD = 3) nurse-coaching sessions during the 90-day post discharge period from the initial hospitalization. In the intervention group, 227 individuals (90%) used the IVR system at least once in the 30-day period following discharge, and 90 individuals (40%) used the system daily for the entire 30-day period. Of the 227 participants who used the IVR at least once over the 30-day post discharge period, 72 (32%) experienced at least 1 acute event, with a total of 92 acute events (readmissions or ED visits). For 59 events (64%), participants triggered a symptom alert on same day as the acute event, and, for 53 events (58%), they triggered an alert on the prior day. For 13 events (14%), participants triggered an alert on both days.
We found no statistically significant differences in either 30-day or 90-day readmission incidence rates (). The 30-day readmission rates were 0.23 per person for the intervention group and 0.19 per person for the usual care group (difference, 0.04; 95% CI, —0.05 to 0.13; P = .36). The 90-day readmission incidence rates were 0.51 and 0.48 per person for the intervention and usual care groups, respectively (difference, 0.04; 95% CI, —0.12 to 0.19; P = .66). Similar results were found when the data were analyzed separately by condition (Table 3), and when adjusted for baseline covariates ().
Overall, as well as by condition, there was no statistically significant difference in ED visit incidence rates between intervention and usual care groups. For the entire sample, the 30-day ED visit incidence rate was 0.17 per person in the intervention group and 0.15 per person in the usual care group (difference, 0.03; 95% CI, —0.05 to 0.11; P = .50). The 90-day ED visit incidence rates were 0.36 and 0.31 per person for intervention and usual care groups, respectively (difference, 0.05; 95% CI, —0.07 to 0.18; P = .41). Results were similar for CHF and COPD, when analyzed separately (Table 3), and when adjusted for baseline covariates (eAppendix C). Similarly, there was no statistically significant difference between groups in the composite rate, whether assessed in aggregate or separately by condition (Table 3).
In the overall sample, 14 (5.5%) participants in the intervention group died within 90 days compared with 26 (10.0%) of participants in the usual care group (). While not statistically significant (P = .06), this difference was primarily driven by COPD participants, and among these participants, the 90-day mortality rate was significantly lower in the intervention group compared with the usual care group. Four (3.2%) individuals in the intervention group died by 90 days, compared with 15 (11.5%) receiving usual care (P = .81; risk ratio [RR] = 0.90, 95% CI 0.40-2.05; risk difference = —0.01, 95% CI –0.08 to 0.06). Among CHF participants, there was no statistically significant difference between groups in the rates of death at 90 days. Ten (7.8%) participants in the intervention group died by 90 days compared with 11 (8.6%) of those receiving usual care (P = .81; RR = 0.90, 95% CI 0.40-2.05; risk difference = —0.01, 95% CI –0.08 to 0.06). Results were similar when mortality rates were adjusted for baseline covariates ().
To inform the design of hospital efforts to avoid financial penalties for readmissions, we tested a comprehensive set of transitional care activities under a hospital’s control, 4,6 and added 2 additional components—MI-based health coaching and symptom monitoring using IVR. Despite the robustness of the intervention, it failed to reduce readmissions or ED visits for participants hospitalized for CHF or COPD.
There are several potential explanations for the lack of effect. The first is the possibility that the intervention was poorly executed. While IVR adherence was below 100%, engagement rates were similar to those reported in Chaudhry et al27 that utilized the same IVR system in their randomized controlled trial of CHF patients (which failed to find a reduction in all-cause 180 day readmissions). Similarly, the quality of MI was at levels above the proficiency threshold, and MI has been found to be an effective intervention for the complex conditions found in healthcare settings.28 Moreover, PAM scores, a measure of patient activation and a marker for patient self-management, showed significantly greater increases among the intervention participants over the 90-day post discharge period (data available from the authors).
A second explanation is that the study population was sufficiently ill to the extent that readmissions were not avoidable. Indeed, a recent study indicated that only 10% of the acute care costs incurred in a cohort of high-cost Medicare patients, with clinical characteristics similar to those of our study population, were considered preventable. 29 However, other studies have successfully reduced readmissions in chronically ill individuals using many of the same intervention components as those used in the current study.22-24,30
A third explanation is that the intervention components that we were not able to include—home visits and ambulatory follow-up care—were essential. That is, even increasing self-management skills and patient activation among intervention participants may not have been sufficient to overcome shortcomings in patients’ clinical management after they were discharged. While home visits might have helped reduce readmissions, only 3 of 9 studies reviewed by Hansen et al4 that included such visits found statistically significant reductions in readmissions. Given that our study participants had ready access to the study nurses, the lack of home visits is unlikely to explain the null finding.
On the other hand, evidence does suggest that patients receiving early follow-up ambulatory care after discharge have lower rates of all-cause 30-day readmission.31,32 Study nurses did send letters to the PCPs of the intervention participants after discharge, call clinic staff to assist with follow-up appointments, and encourage participants to visit their PCP when experiencing a change in symptoms or health status. However, as our study did not include the active or formal participation of community-based providers, these activities may not have been sufficient to ensure robust provider continuity and timely clinic follow-up. We have reason to believe this to be the primary factor limiting the effectiveness of the intervention. In an extensive review of intervention group cases with the nurses at the completion of the study, in all but 1 case, the patients’ providers either declined to schedule a same-day appointment or instructed the study nurses to send the patients directly to the ED. In the single case where the physician did schedule a same-day visit, the acute exacerbation was resolved in the outpatient setting. Thus, there is anecdotal evidence to suggest that in our study, early and effective primary medical care could have prevented at least some readmissions.
Taken together, there is a real possibility that standalone community hospitals may have limited impact on the prevention of readmission for severely ill patients with COPD or CHF, despite the use of comprehensive, evidence-based interventions that are within their control Nonetheless, instituting programs that increase patient engagement in self-management may still be beneficial, and the reduction in mortality among COPD patients in our study is particularly encouraging. Improved self-management may have been beneficial to COPD patients in our study because of the nature of the self-care regimens. Vestbo et al33 found that adherence to inhaled medication was significantly associated with reduced risk of death due to exacerbations in COPD, and it appears that this effect was nearly immediate. Our study nurses reported that, at enrollment, many COPD participants had either been nonadherent to their inhaled medications or had been using inhalers incorrectly. Intervention participants received substantial instruction on the correct use of these medications, and the MI sessions focused on evoking the benefits of the treatment plan. This may explain the reduction in mortality among COPD patients that we found.
Our study had multiple strengths. First, we had 512 participants, which is larger than the 73% of the randomized controlled trials reported by Hansen et al.4 Second, we selected 2 medical conditions that have high readmission prevalence and are the focus of Medicare’s policy to penalize hospitals for 30-day readmissions. Third, our robust intervention included MI and IVR, 2 separately effective interventions that heretofore have not been included, or studied, as integral components of the transitional care model. And fourth, our study was conducted in 2 stand-alone community hospitals, rather than in an academic hospital or integrated health system where most clinical trials are typically conducted.
Our study also had limitations. We did not have access to data on readmissions or ED visits that may have occurred at other facilities, beyond self-reported data from participants. However, we can presume any missed data to be evenly distributed between intervention and control groups due to randomization. Additionally, our study was conducted at 2 hospitals in southern Oregon, which may limit the generalizability of the results to other areas of the country, although no specific factors emerged suggesting unique geographic influences on the intervention. Finally, as described above, we did not include home visits or have the ability to ensure timely follow-up ambulatory care.
In summary, a comprehensive hospital-based intervention failed to reduce 30- or 90-day readmissions as well as ED visits for patients with CHF or COPD, compared with usual care. It did, however, reduce 90-day mortality among COPD patients. Our results suggest the need to continue to experiment with new interventions targeting readmissions—in particular, those focused on building collaborative relationships between hospitals and community-based providers. In the interim, our results point to a challenging road ahead for hospitals seeking to decrease readmissions for chronically ill patients and avoid financial penalties.
This study would not be possible without the dedication of the project team at Asante Health System. We thank Debra Flickinger, BSM, accreditation and quality control coordinator; Debbie Daggett, BA, grant manager; Sandra Olson, BS, director of research and development; Kimberly Swegar, RN, MSN, study nurse; Dion Marchi, RN, BSN, study nurse; and Christal Paetz, BA, clinical informatics liaison, who were compensated for their work. We are grateful to the Health Resources and Services Administration for providing the funding to support this trial. Lastly, we thank Julia Adler-Milstein, PhD, Nancy Adler, PhD, and Ashish Jha, MD, for their critical review and editing of earlier drafts of this manuscript, for which they received no compensation. Finally, we would like to thank the editor and anonymous reviewers for providing instructive feedback that improved the manuscript.Author Affiliations: From Linden Consulting Group, LLC, Ann Arbor, MI (AL); University of Michigan, School of Public Health, Department of Health Management & Policy, Ann Arbor (AL); Q-consult, LLC, Portland, OR (SWB); and Oregon Health & Science University, Portland (SWB).
Source of Funding: This study was funded by grant H2AIT16621 from the Health Resources and Services Administration (HRSA). However, the HRSA did not have a role in the design and conduct of the study; in the collection, management, analysis, or interpretation of the data; or in the preparation, review, or approval of the manuscript.
Author Disclosures: The authors report no conflicts of interest.
Authorship Information: Concept and design (AL, SWB); acquisition of data (AL); analysis and interpretation of data (AL, SWB); drafting of the manuscript (AL, SWB); critical revision of the manuscript for important intellectual content (AL, SWB); statistical analysis (AL); obtaining funding (AL); administrative, technical, or logistic support (AL, SWB); supervision (AL, SWB).
Address correspondence to: Ariel Linden, DrPH, Linden Consulting Group, LLC, 1301 N Bay Dr, Ann Arbor, MI 48103. E-mail: email@example.com.REFERENCES
1. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med. 2009;360(14):1418-1428.
2. Patient Protection and Affordable Care Act (PPACA). Public Law 111—148; 2010. http://www.gpo.gov/fdsys/pkg/PLAW- 111publ148/pdf/PLAW-111publ148.pdf. Accessed on May 30, 2013.
3. National Committee for Quality Assurance (NCQA). 2012 insights for improvement — reducing readmissions: measuring health plan performance. Washington, DC: National Committee for Quality Assurance; 2012. http://www.ncqa.org/portals/0/Publications/2012%20BI_NCQA%20ReAdMi%20_Pub.pdf. Accessed June 2, 2013.
4. Hansen LO, Young RS, Hinami K, Leung A, Williams MV. Interventions to reduce 30-day rehospitalization: a systematic review. Ann Intern Med. 2011;155(8):520-528.
5. Ryan P. Random allocation of treatments in blocks. Stata Technical Bulletin. 1998;41(1):43-46. http://ideas.repec.org/c/boc/bocode/s319901.html. Accessed April 21, 2013.
6. Coleman EA, Berenson RA. Lost in transition: challenges and opportunities for improving the quality of transitional care. Ann Intern Med. 2004;141(7):533-536.
7. Miller WR, Rollnick S. Motivational Interviewing: Helping People Change. 3rd ed. New York, NY: Guilford Press; 2012.
8. Lundahl B, Burke BL. The effectiveness and applicability of motivational interviewing: a practice-friendly review of four meta-analyses. J Clin Psychol. 2009;65(11):1232-1245.
9. Benzo R, Vickers K, Ernst D, et al. Development and feasibility of a self-management intervention for chronic obstructive pulmonary disease delivered with motivational interviewing strategies. J Cardiopulm Rehabil Prev. 2013;33x:113-122.
10. Linden A, Butterworth SW, Prochaska JO. Motivational interviewing-based health coaching as a chronic care intervention. J Eval Clinical Pract. 2010;16(1):166-174.
11. Lundahl BW, Kunz C, Brownell C, et al. A meta-analysis of motivational interviewing: twenty-five years of empirical studies. Res Work Soc Pract. 2010;20(2):137-160.
12. Brodie DA, Inoue A, Shaw DG. Motivational interviewing to change quality of life for people with chronic heart failure: a randomised controlled trial. Int J Nurs Stud. 2008;45(4):489-500.
13. Inglis SC, Clark RA, McAlister FA, et al. Structured telephone support or telemonitoring programmes for patients with chronic heart failure. Cochrane Database Syst Rev. 2010;8:CD007228.
14. Polisena J, Tran K, Cimon K, et al. Home telehealth for chronic obstructive pulmonary disease: a systematic review and meta-analysis. J Telemed Telecare. 2010;16(3):120-127.
15. Moyers TB, Martin T, Manuel JK, Miller WR, Ernst D. Revised Global Scales: Motivational Interviewing Treatment Integrity 3.1.1. Albuquerque, NM: University of New Mexico Center on Alcoholism, Substance Abuse, and Addictions; 2010. http://casaa.unm.edu/download/miti3_1.pdf. Accessed October 1, 2014.
16. Connolly M, Grimshaw J, Dodd M, et al. Systems and people under pressure: the discharge process in an acute hospital. J Clin Nurs. 2009;18(4):549-558.
17. Rabe KF, Hurd S, Anzueto A, et al; Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: GOLD executive summary. Am J Respir Crit Care Med. 2007;176(6):532-555.
18. Hibbard JH, Mahoney ER, Stockard J, Tusler M. Development and testing of a short form of the patient activation measure. Health Serv Res. 2005;40(6, pt 1):1918-1930.
19. Hibbard JH, Greene J. What the evidence shows about patient activation: better health outcomes and care experiences; fewer data on costs. Health Aff. 2013;32(2):207-214.
20. Newson R. Generalized power calculations for generalized linear models and more. Stata J. 2004;4(4):379-401.
21. Naylor M, Brooten D, Jones R, Lavizzo-Mourey R, Mezey M, Pauly M. Comprehensive discharge planning for the hospitalized elderly. a randomized clinical trial. Ann Intern Med. 1994;120(12):999-1006.
22. Naylor MD, Brooten D, Campbell R, et al. Comprehensive discharge planning and home follow-up of hospitalized elders: a randomized clinical trial. JAMA. 1999;281(7):613-620.
23. Benatar D, Bondmass M, Ghitelman J, Avitall B. Outcomes of chronic heart failure. Arch Intern Med. 2003;163(3):347-352.
24. Weintraub AR, Kimmelstiel C, Levine D, et al. A multicenter randomized controlled comparison of telephonic disease management vs. automated home monitoring in patients recently hospitalization with heart failure: Span-CHF II Trial. J Card Fail. 2005;11(9):720.
25. Long JS, Freese J. Regression Models for Categorical Dependent Variables Using Stata. 2nd ed. College Station, TX: Stata Press; 2006.
26. Norton EC, Miller MM, Kleinman LC. Computing adjusted risk ratios and risk differences in Stata. Stata J. 2013;13(3);492-509.
27. Chaudhry SI, Mattera JA, Curtis JP, et al. Telemonitoring in patients with heart failure. N Engl J Med. 2010;363(24):2301-2309.
28. Lundahl B, Moleni T, Burke BL, et al. Motivational interviewing in medical care settings: a systematic review and meta-analysis of randomized controlled trials. Patient Educ Couns. 2013;93(2):157-168.
29. Joynt KE, Gawande AA, Orav EJ, Jha AK. Contribution of preventable acute care spending to total spending for high-cost Medicare patients. JAMA. 2013;309(24):2572-2578.
30. Coleman EA, Parry C, Chalmers S, Min SJ. The care transitions intervention: results of a randomized controlled trial. Arch Intern Med. 2006;166(17):1822-1828.
31. Hernandez AF, Greiner MA, Fonarow GC, et al. Relationship between early physician follow-up and 30-day readmission among Medicare beneficiaries hospitalized for heart failure. JAMA. 2010;303(17):1716-1722.
32. Misky GJ, Wald HL, Coleman EA. Post-hospitalization transitions: examining the effects of timing of primary care provider follow-up. J Hosp Med. 2010;5(7):392-397.
33. Vestbo J, Anderson JA, Calverley PM, et al. Adherence to inhaled therapy, mortality and hospital admission in COPD. Thorax. 2009;64(11):939-943.