Does Paid Versus Unpaid Supplementary Caregiving Matter in Preventable Readmissions?

March 17, 2017

Home health beneficiaries with diabetes using paid supplementary caregivers had 68% higher hazards of readmission due to urinary tract infection than those with unpaid supplementary caregivers.

ABSTRACTObjectives: To examine the impact of paid and unpaid supplementary caregiving on preventable readmissions among Medicare home health beneficiaries with diabetes.

Study Design: A retrospective cohort study with a home health episode as the unit of analysis.

Methods: We gathered multiple 2009 Medicare data and national datasets (eg, Area Health Resources File). We used propensity score matching based on beneficiaries’ predisposing, enabling, and health need factors to create matched cohorts for episodes solely assisted by paid supplementary caregivers versus those solely assisted by unpaid supplementary caregivers. We applied Cox regression on the matched cohorts to estimate the 30-day preventable readmissions, including diabetes-related conditions, heart disease-related conditions, chronic obstructive pulmonary disease, urinary tract infection (UTI), pneumonia, and dehydration.

Results: Among beneficiaries (81,481) who received assistance several times during the day and night or several times during only the day, 7.34% were solely assisted by paid supplementary caregivers and 86.34% were solely assisted by unpaid supplementary caregivers. In the matched cohorts, we found that beneficiaries with paid supplementary caregivers had 68% higher hazards of readmission due to UTIs than those with unpaid supplementary caregivers.

Conclusions: The Medicare program is moving toward value-based purchasing and penalizing home health agencies with poor quality of care. It is critical for home health professionals to invite paid and unpaid supplementary caregivers to initiate care plans and assess their competence. Policies that support unpaid supplementary caregivers can also assist caregivers to care for their loved ones and prevent them from using expensive hospital resources.

Am J Manag Care. 2017;23(3):e82-e88

Takeaway Points

  • Among beneficiaries who received assistance several times during the day and night or several times during only the day, 7.34% and 86.34% were assisted solely by paid and unpaid supplementary caregivers, respectively.
  • Beneficiaries with diabetes and paid supplementary caregivers had a 68% higher hazard of readmission due to urinary tract infections than those with unpaid supplementary caregivers.
  • The Medicare program is moving toward value-based purchasing. It is critical for home health professionals to invite paid and unpaid supplementary caregivers to initiate care plans and assess their competence. Policies supporting unpaid supplementary caregivers can reduce beneficiaries’ used of expensive hospital resources.

Approximately 3.4 million Medicare home health beneficiaries receive professional home health services that are paid for by Medicare.1 These beneficiaries, on average, received 19 visits from home health professionals during an episode of home healthcare, defined as up to 60 days, in 2014.1 A study based on a large home health agency in New York found that in addition to receiving services from home health professionals, 83.4% of home health beneficiaries received supplementary care from informal caregivers, of which 7% were paid, in order to meet their needs.2 These paid informal caregivers (hereafter, called paid supplementary caregivers) are paid through beneficiaries and/or families out of pocket or through other programs (eg, Medicaid).

The interpersonal relationships between home health beneficiaries and unpaid informal caregivers (hereafter, called unpaid supplementary caregivers) and the interpersonal relationships between home health beneficiaries and paid supplementary caregivers are different. Unpaid supplementary caregivers are beneficiaries’ family members, relatives, and/or friends. They have solid relationships with their patients and understand what patients need without requiring much communication whereas paid supplementary caregivers are introduced by home health agencies and have no prior relationships with the beneficiaries and/or their families. Therefore, they need to learn how to meet patients’ needs through communication that is often difficult, given a high prevalence of limitation in cognitive function among Medicare home health beneficiaries.3

Regardless of the differences in the interpersonal relationships, paid and unpaid supplementary caregivers provide emotional and social support and a wide scope of direct care to patients, ranging from necessary medical care (eg, medication dispensing and wound care) to assistance with activities of daily living (ADL) (eg, toileting and eating) and/or instrumental activities of daily living (eg, shopping and laundry).4,5 Given the scope of care activities, paid and unpaid supplementary caregivers are a vital part of a healthcare team and have a significant impact on home health beneficiaries’ health outcomes.

About 1.8 million Medicare home health beneficiaries received postacute care in 2012.1 In order to reduce readmission under the Affordable Care Act (ACA), the Medicare Payment Advisory Commission endorsed a home health financial incentive program for reducing preventable readmissions for home health agencies.1 Paid and unpaid supplementary caregivers play a critical role in caring for beneficiaries during the transition from hospitals to homes and postacute care,6 and can therefore impact the likelihood of preventable readmissions. Under a home health financial penalty program, investigating the potential difference in preventable readmissions of Medicare home health beneficiaries with these 2 types of supplementary caregivers is imperative for the home health industry.

The purpose of this study was to compare the likelihood of 30-day preventable readmissions among Medicare home health beneficiaries with diabetes who received care from unpaid supplementary caregivers with that of those who received care from paid supplementary caregivers. We chose diabetes because it is the most common disease (one-third) among Medicare home health beneficiaries,7 and diabetes significantly contributes to disabilities, such as blindness and amputation,8 that are associated with high demands for caregiving.

Conceptual Framework

We used Andersen’s Behavior Model of Health Services Use as a guiding framework for our statistical modeling.9 The Andersen model stated that healthcare utilizations and outcomes are a function of patients’ predisposing (eg, gender, race/ethnicity), enabling (eg, living alone), and health need factors (eg, ADL, cognitive disorders), as well as the characteristics of healthcare organizations and communities. This model fits our study well because evidence shows that the decision of whether to hire a paid caregiver depends on beneficiaries’ predisposing, enabling, and health need factors.10,11 Additionally, supplementary caregivers—one of the beneficiaries’ enabling factors—are part of the healthcare delivery team12 likely to affect beneficiaries’ health outcomes (eg, preventable readmissions).2,13 Lastly, the characteristics of organizations and communities are associated with beneficiaries’ health outcomes.14-16

We applied Andersen’s framework in 2 stages: 1) we used beneficiaries’ predisposing, enabling, and health need factors to estimate the propensity of beneficiaries solely assisted by paid supplementary caregivers and those solely assisted by unpaid supplementary caregivers to construct a matched cohort10,11; and 2) we applied Cox regression on the matched cohort to estimate the difference in time to first 30-day preventable readmissions between these 2 groups, after controlling for the covariates. The conceptual framework is presented in the Figure.

METHODS

Data Sources

Based on the conceptual framework, we utilized several 2009 national datasets in the study. We extracted variables for beneficiary’s predisposing, enabling, and health need factors from several Medicare claim and assessment files discussed in our previous studies.17,18 (Operational definitions are provided in the eAppendix for review; eAppendices are available at ajmc.com.) The American Hospital Association (AHA) Annual Survey and Provider of Services (POS) file provide the characteristics (eg, ownership) of hospitals and home health agencies, respectively. Finally, the Area Health Resources File (AHRF) and Primary Care Service Area (PCSA) database from HHS provide the characteristics of communities where patients reside.

Study Design and Sample

We conducted a retrospective cohort study with home health episode for postacute care as the unit of analysis. We identified episodes of postacute care—defined as receiving home health services within 14 days of hospital discharge1,16,19—by merging the Medicare Provider Analysis and Review (MedPAR) that provides the date of hospital discharge and the Outcome Assessment Information Set (OASIS) that provides the start date of home healthcare, recommended by the literature.19

In order to achieve our study aim, we used several criteria to select our study sample. First, beneficiaries must have received assistance several times during the day and night or several times only during the day. The OASIS B1 version provides the frequency of assistance (M0370),20 which allowed us to exclude beneficiaries who received assistance once daily or less, whose health outcomes are less likely to be significantly affected by supplementary caregivers. Second, beneficiaries could not be dual eligible because their family members could be paid supplementary caregivers under the Medicaid program. As discussed previously about interpersonal relationships, these family members as unpaid supplementary caregivers are quite different from the paid supplementary caregivers who are paid out of pocket by the beneficiaries and/or their families and do not have prior relationships to the beneficiaries and their families. We used the Master Beneficiary Summary File to determine dual eligibility and excluded these individuals from our study sample. Third, beneficiaries must solely rely on paid or unpaid supplementary caregivers (we provided detailed explanation in the subsection of “Paid and Unpaid Supplementary Caregivers”). Finally, beneficiaries must have been 65 years or older and enrolled in the fee-for-service (FFS) program.

Study Variables

Paid and unpaid supplementary caregivers. We used 2 assessment codes from the OASIS B1 version to identify beneficiaries with solely paid or unpaid supplementary caregivers. The assessment code M0360 provides information regarding which primary caregiver is taking the lead responsibility.21 The categories of primary informal caregivers in code M0360 include a spouse or significant other, daughter or son, other family member, friend, neighbor, member of the community or church members, or paid helper.21 The assessment code M0350A-M0350E (M0350A indicates that beneficiaries did not have an assisting individual) provides information on whether assisting individuals were paid (M0350B) or unpaid helpers (M0350C-M0350E).22

Beneficiaries with solely paid supplementary caregivers were beneficiaries whose code for M0360 was paid helper and M0350B was coded as 1; namely, beneficiaries had paid helpers as primary informal caregivers and received assistance from paid helpers. Beneficiaries with solely unpaid supplementary caregivers were beneficiaries whose code for M0360 was unpaid helper (ie, a spouse or significant other, daughter or son, other family member, friend, neighbor, member of the community, or church members) and the code for M0350C, M0350D, or M0350E was 1; namely, beneficiaries had unpaid helpers as primary informal caregivers and received assistance from these individuals. Patients who had unpaid supplementary caregivers as primary caregivers and received assistance from paid helpers or vice versa were excluded because we are unable to determine the proportion of care provided by paid and unpaid supplementary caregivers.

Variables for matching. After identifying the qualified study sample, we selected a series of variables from beneficiaries’ predisposing, enabling, and health need factors and applied the propensity score matching to match beneficiaries with paid supplementary caregivers and those with unpaid supplementary caregivers.23,24 The predisposing factors included age, race/ethnicity, and gender. The enabling factors included whether or not beneficiaries lived alone. The heath need factors included ADL limitations, cognitive disorders, a presence of pressure or stasis ulcer, whether or not beneficiaries felt anxious or used a urinary catheter, and a dummy variable for beneficiaries with 3 or more comorbidities defined by Elixhauser and colleagues.25 The operational definition of the variables for matching was discussed in our previous studies (also presented in the eAppendix).

Dependent variables. The dependent variable was 30-day preventable readmissions. We applied the Agency for Health Research and Quality (AHRQ) Prevention Quality Indicator (PQI) software version 4.5 on the MedPAR to identify readmissions due to preventable conditions.26 These conditions include diabetes with short-term complications (PQI 01), perforated appendix (PQI 02), diabetes with long-term complications (PQI 03), chronic obstructive pulmonary disease (COPD) or asthma in older adults (PQI 05), hypertension (PQI 07), heart failure (PQI 08), dehydration (PQI 10), bacterial pneumonia (PQI 11), urinary tract infection (UTI) (PQI 12), angina without procedure (PQI 13), uncontrolled diabetes (PQI 14), and lower-extremity amputation among patients with diabetes (PQI 16). Hospitalizations due to these conditions are preventable if patients are able to receive proper and timely care in the community.26-28 Home health beneficiaries receive professional home healthcare at their residences; with proper care and interventions after hospital discharge, readmissions due to preventable conditions identified by AHRQ are avoidable for our study population.

We excluded perforated appendix because none of the matched cohorts had readmission due to this condition. Because the frequency for some preventable conditions was small (eg, hypertension), we regrouped the preventable conditions based on the human organ system. The groups for preventable readmissions are: 1) diabetes-related conditions (PQI 01, PQI 03, PQI 14, and PQI 16), 2) heart disease-related conditions (PQI 07, PQI 08, PQI 13), 3) COPD, 4) UTI, 5) pneumonia, and 6) dehydration. These preventable readmissions were finally defined as a time-to-first event in our analytical model.

Covariates for estimated model. Given the conceptual framework, the covariates in the analytical model are the intensity of home health visits and the characteristics of hospitals, home health agencies, and communities. Hospital characteristics—extracted from the AHA Annual Survey—included hospitals’ ownership, teaching status, and system affiliation. Ownership of home health agencies was extracted from the POS files. Community characteristics, which were extracted from the AHRF, included a rural/nonrural county characterization and the number of primary care physicians and acute hospital beds per 1000 population at the county level. A log-transformed median household income at the zip code level was extracted from the PCSA. (Information regarding how to merge all datasets is available in the eAppendix.)

Statistical Methods

Matching model to find matched cohorts. Approximately 4% of beneficiaries with diabetes had more than 1 home health episode for post acute care in 2009. To account for the cluster from the same beneficiary using multiple home health episodes, we applied a logistic regression with the Huber Sandwich Estimator on the selected variables from patients’ predisposing, enabling, and health need factors to estimate the propensity score. We then applied a caliper without replacement on propensity score to find the matched cohorts.23,24 We compared the 3 factors of beneficiaries solely assisted by paid supplementary caregivers versus those solely assisted by unpaid supplementary caregivers, before and after matching, using χ2 testing.

Estimated model on matched cohorts. We applied a Cox proportional hazards regression model on the matched cohorts to estimate the adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for paid supplementary caregivers, after adjusting for the covariates discussed previously. The Cox regression model was applied to 6 preventable conditions individually. Approximately 1% of the matched cohorts were from the same beneficiary; thus, we fit the model with standard errors adjusted for clustering on each individual beneficiary.

RESULTS

We identified 120,208 diabetes-related postacute home health episodes. Among them, 85.79% had primary caregivers (based on code M0360 in the OASIS) and 98% received assistance from paid or unpaid caregivers (based on codes M0350A-M0350E). After excluding unqualified episodes (previously discussed in the study sample), we identified 81,481 episodes in the unmatched study sample, with 70,350 (86.34%) and 5982 (7.34%) observations having solely unpaid and solely paid supplementary caregivers, respectively (5149 observations had both paid and unpaid supplementary caregivers). Before matching, patients’ predisposing, enabling, and health need factors statistically differed between beneficiaries with paid supplementary caregivers and those with unpaid supplementary caregivers (P <.05). In our matched sample (5860 observations for each group), the difference in predisposing, enabling, and health need factors between these 2 groups became statistically insignificant. Table 1 presents the descriptive analysis for these 3 matching factors between unpaid and paid supplementary caregivers before and after matching.

Table 2 presents the descriptive analysis for the dependent variables and covariates for the matched cohorts. Among readmission conditions, beneficiaries solely assisted by paid supplementary caregivers had about 0.46% higher readmissions due to UTIs than those solely assisted by unpaid supplementary caregivers (P <.01). Table 3 presents the HR and 95% CI from the Cox regression models for the matched cohorts. For UTIs, the HR is 1.68 (95% CI, 1.13-2.50); namely, beneficiaries with solely paid supplementary caregivers had 68% higher hazards of readmission due to UTIs than whose with solely unpaid supplementary caregivers. We did not find statistical differences for COPD, heart-related diseases, dehydration, diabetes-related conditions, and pneumonia.

DISCUSSION

Although existing literature primarily focuses on volunteer family caregivers, there is a growing interest in studying paid informal caregivers due to the reduction in family size and rising number of women in the workforce. Lindquist et al (2011, 2012) found that paid informal caregivers had a good heart to help elderly patients, but were likely to have low health literacy and receive little training before they were assigned to patients’ homes.29,30 Among previous studies, one focused on Medicare home health beneficiaries found that beneficiaries with paid informal caregivers had poorer functional improvement than those with unpaid informal caregivers.2 Another study found that Medicare heart failure beneficiaries with paid informal caregivers had a higher risk of 1-year readmission than those with unpaid informal caregivers.13

Among all preventable readmissions, we found that beneficiaries solely assisted by paid supplementary caregivers had a significantly higher hazard of readmissions due to UTIs than those solely assisted by unpaid supplementary caregivers. Given that prevention of a UTI requires good genital care and regular bladder emptying, this type of personal care necessitates a good relationship between patients and supplementary caregivers. Paid supplementary caregivers in our study were not beneficiaries’ family members; one study showed that patients with family caregivers felt safer and had higher satisfaction in their interpersonal relationship than those with paid nonfamily caregivers,31 which explains our significant findings for readmissions due to UTI.

A previous study based on the information from code M0360 in the OASIS found that 83.4% of Medicare home health beneficiaries had primary informal caregivers.2 Based on code M0360, our study found that 85.79% of Medicare home health beneficiaries with diabetes had primary informal caregivers. Our study moved further and found that for the beneficiaries who received assistance several times during the day and night or several times during only the day, 7.34% were solely assisted by paid supplementary caregivers, 86.34% were solely assisted by unpaid supplementary caregivers, and 6.32% received assistance from both paid and unpaid supplementary caregivers. Our findings also showed that paid and unpaid supplementary caregivers heavily involved in caregiving had a significant impact on health outcomes for these frail, community-dwelling homebound beneficiaries. In order to obtain the full picture for the difference in quality of care provided by paid and unpaid supplementary caregivers, future studies focusing on other quality indicators, such as emergency visits and mortality, are recommended.

Limitations

Our datasets did not provide information on competence, stress, and support systems for paid and unpaid supplementary caregivers, which are likely to affect patients’ outcomes.32 Additionally, we focused on Medicare FFS home health beneficiaries with diabetes without identifying type 1 and 2 diabetes or insulin users. We also excluded beneficiaries with dual eligibility, those receiving assistance once daily or less, and those receiving care from both paid and unpaid supplementary caregivers. Furthermore, we used OASIS, which contains only data in the United States, to provide clinical and functional conditions for Medicare home health beneficiaries and allows researchers to merge with other Medicare claim data, such as MedPAR. Readers must also be aware of the debate concerning the reliability and validity of OASIS.33,34

Moreover, we used preventable conditions defined by AHRQ. The conditions for preventable readmissions defined in the 3M Potentially Preventable Readmissions Grouping Software are slightly different from the preventable conditions defined by AHRQ. Future studies based on the 3M Potentially Preventable Readmissions Grouping Software are encouraged. Also, the propensity score matching was based on the variables available in our data. There were missing variables in our study that should be taken into consideration when interpreting our findings. Our datasets do not have beneficiaries’ social/contextual factors, such as their education and whether they received social support from other organizations that were associated with the likelihood of hiring a supplementary caregiver and preventable readmissions.

Finally, we used 2009 datasets. Several programs, such as the Hospital Readmissions Reduction Program introduced in 2013 under the ACA,35 are likely to have a spillover effect on preventable readmissions for Medicare home health beneficiaries. Future studies using recent data are recommended.

CONCLUSIONS

Despite the study limitations, our findings have clinical and policy implications. For clinical implications, paid and unpaid supplementary caregivers are critical partners for home health professionals. Inviting supplementary caregivers to initiate care plans and assuring their competence to meet Medicare home health beneficiaries’ health needs becomes critical in the home health industry as the Medicare program is moving toward value-based purchasing and penalizing home health agencies with high preventable readmissions. For policy implications, our findings showed that our society heavily relies on unpaid supplementary caregivers to provide care to meet their loved ones’ health needs and that beneficiaries with unpaid supplementary caregivers had lower risks of readmission due to UTIs than those with paid supplementary caregivers. Policies that support unpaid supplementary caregivers are critical to help caregivers assist their loved ones and prevent them from using expensive hospital resources.36,37

Acknowledgments

The authors thank Dr Sharon Homan’s contribution in the early stage of this project.

Author Affiliations: Department of Health Policy and Management (HFC), College of Public Health, University of Arkansas for Medical Sciences, Little Rock, AR; Department of Biostatistics and Epidemiology, School of Public Health (SS), University of North Texas Health Science Center, Fort Worth, TX; Department of Health Policy and Management, School of Medicine, University of Kansas Medical Center (TP), Kansas City, KS.

Source of Funding: This study was funded by the Junior Faculty Seed Fund at the University of North Texas Health Science Center (UNTHSC). The views expressed in this publication are the views of the authors and do not necessarily reflect the views of UNTHSC.

Author Disclosures: The manuscript was accepted when Dr Chen was affiliated with UNTHSC and was published when Dr Chen was affiliated with the University of Arkansas for Medical Sciences. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (HFC); acquisition of data (HFC); analysis and interpretation of data (HFC, SS); drafting of the manuscript (HFC, TP, SS); critical revision of the manuscript for important intellectual content (HFC, TP, SS); statistical analysis (HFC, TP); obtaining funding (HFC); and administrative, technical, or logistic support (HFC).

Address Correspondence to: Hsueh-Fen Chen, PhD, Department of Health Policy and Management, University of Arkansas for Medical Sciences 4301 West Markham St, Mail Slot 820-12 Little Rock, AR 72205. E-mail: hchen@uams.edu.

REFERENCES

1. Report to the Congress: Medicare payment policy—chapter 9: home health services. Medicare Payment Advisory Commission website. http://www.medpac.gov/docs/default-source/reports/mar2015_entirereport_revised.pdf?sfvrsn=0. Published March 2015. Accessed February 3, 2017.

2. Cho E, Kim EY, Lee NJ. Effects of informal caregivers on function of older adults in home health care. West J Nurs Res. 2013;35(1):57-75. doi: 10.1177/0193945911402847.

3. Avalere Health LLC. The Medicare home health benefit/program costs and beneficiary characteristics. Almost Family website. http://www.almostfamily.com/pdf/20090715_AHHQI%20Chart%20Book.pdf. Published July 15, 2009. Accessed February 3, 2017.

4. Kaye HS, Harrington C, LaPlante MP. Long-term care: who gets it, who provides it, who pays, and how much? Health Aff (Millwood). 2010;29(1):11-21. doi: 10.1377/hlthaff.2009.0535.

5. Donelan K, Hill CA, Hoffman C, et al. Challenged to care: informal caregivers in a changing health system. Health Aff (Millwood). 2002;21(4):222-231.

6. Naylor MD, Brooten DA, Campbell RL, Maislin G, McCauley KM, Schwartz JS. Transitional care of older adults hospitalized with heart failure: a randomized, controlled trial. J Am Geriatr Soc. 2004;52(5):675-684.

7. Caffrey C, Sengupta M, Moss A, Harris-Kojetin L, Valverde R. Home health care and discharged hospice care patients: United States, 2000 and 2007. Natl Health Stat Report. 2011;27(38):1-27.

8. National diabetes fact sheet: national estimates and general information on diabetes and pre-diabetes in the United States, 2011. CDC website. http://www.cdc.gov/diabetes/pubs/pdf/ndfs_2011.pdf. Published 2011. Accessed January 5, 2016.

9. Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav. 1995;36(1):1-10.

10. DeFries EL, McGuire LC, Andresen EM, Brumback BA, Anderson LA. Caregivers of older adults with cognitive impairment. Prev Chronic Dis. 2009;6(2):A46.

11. LaPlante MP, Harrington C, Kang T. Estimating paid and unpaid hours of personal assistance services in activities of daily living provided to adults living at home. Health Serv Res. 2002;37(2):397-415.

12. Institute of Medicine of the National Academies. Retooling for an aging America: building the health care workforce: report brief. The National Academies of Sciences, Engineering, and Medicine website. http://www.nationalacademies.org/hmd/~/media/Files/Report%20Files/2008/Retooling-for-an-Aging-America-Building-the-Health-Care-Workforce/ReportBriefRetoolingforanAgingAmericaBuildingtheHealthCareWorkforce.pdf. Published April 2008. Accessed June 6, 2016.

13. Mosca L, Aggarwal B, Mochari-Greenberger H, et al. Association between having a caregiver and clinical outcomes 1 year after hospitalization for cardiovascular disease. Am J Cardiol. 2012;109(1):135-139. doi: 10.1016/j.amjcard.2011.07.072.

14. Joynt KE, Jha AK. Thirty-day readmissions—truth and consequences. N Engl J Med. 2012;366(15):1366-1369. doi: 10.1056/NEJMp1201598.

15. Herrin J, St Andre J, Kenward K, Joshi MS, Audet AM, Hines SC. Community factors and hospital readmission rates. Health Serv Res. 2015;50(1):20-39. doi: 10.1111/1475-6773.12177.

16. Madigan EA, Gordon NH, Fortinsky RH, Koroukian SM, Piña I, Riggs JS. Rehospitalization in a national population of home health care patients with heart failure. Health Serv Res. 2012;47(6):2316-2338. doi: 10.1111/j.1475-6773.2012.01416.x.

17. Chen HF, Homan S, Carlson E, Popoola T, Radhakrishnan K. The impact of race and neighborhood racial composition on preventable readmissions for diabetic Medicare home health beneficiaries [published online August 11, 2016]. J Racial Ethn Health Disparities.

18. Chen HF, Carlson E, Popoola T, Suzuki S. The impact of rurality on 30-day preventable readmission, illness severity, and risk of mortality for heart failure Medicare home health beneficiaries. J Rural Health. 2016;32(2):176-187. doi: 10.1111/jrh.12142.

19. Wolff JL, Meadow A, Weiss CO, Boyd CM, Leff B. Medicare home health patients’ transitions through acute and post-acute care settings. Med Care. 2008;46(11):1188-1193. doi: 10.1097/MLR.0b013e31817d69d3.

20. (M0370) Frequency patient receives assistance. Research Data Assistance Center website. http://www.resdac.org/cms-data/variables/M0370-Frequency-Patient-Receives-Assistance. Accessed October 25, 2015.

21. (M0360) Primary caregiver. Research Data Assistance Center website. http://www.resdac.org/cms-data/variables/M0360-Primary-Caregiver. Accessed October 25, 2015.

22. Home health outcome and assessment information set. Research Data Assistance Center. website. https://www.resdac.org/cms-data/files/oasis/data-documentation. Accessed June 6, 2016.

23. Rubin DB, Thomas N. Matching using estimated propensity scores: relating theory to practice. Biometrics. 1996;52(1):249-264.

24. Austin PC. Statistical criteria for selecting the optimal number of untreated subjects matched to each treated subject when using many-to-one matching on the propensity score. Am J Epidemiol. 2010;172(9):1092-1097. doi: 10.1093/aje/kwq224.

25. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8-27.

26. AHRQ Quality Indicators—guide to prevention quality indicators: hospital admission for ambulatory care sensitive conditions [AHRQ pub No. 02-R0203]. Agency for Health Research and Quality website. http://www.ahrq.gov/downloads/pub/ahrqqi/pqiguide.pdf. Published October 2001. Accessed December 10, 2015.

27. Billings J, Zeitel L, Lukomnik J, Carey TS, Blank AE, Newman L. Impact of socioeconomic status on hospital use in New York City. Health Aff (Millwood).1993;12(1):162-173.

28. Bindman AB, Grumbach K, Osmond D, et al. Preventable hospitalizations and access to health care. JAMA. 1995;274(4):305-311.

29. Lindquist LA, Jain N, Tam K, Martin GJ, Baker DW. Inadequate health literacy among paid caregivers of seniors. J Gen Intern Med. 2011;26(5):474-479. doi: 10.1007/s11606-010-1596-2.

30. Lindquist LA, Cameron KA, Messerges-Bernstein J, et al. Hiring and screening practices of agencies supplying paid caregivers to older adults. J Am Geriatr Soc. 2012;60(7):1253-1259. doi: 10.1111/j.1532-5415.2012.04047.x.

31. Benjamin AE, Matthias R, Franke TM. Comparing consumer-directed and agency models for providing supportive services at home. Health Serv Res. 2000;35(1 pt 2):351—366.

32. Shankar KN, Hirschman KB, Hanlon AL, Naylor MD. Burden in caregivers of cognitively impaired elderly adults at time of hospitalization: a cross-sectional analysis. J Am Geriatr Soc. 2014;62(2):276-284. doi: 10.1111/jgs.12657.

33. O’Connor M, Davitt JK. The Outcome and Assessment Information Set (OASIS): a review of validity and reliability. Home Health Care Serv Q. 2012;31(4):267-301. doi: 10.1080/01621424.2012.703908.

34. Madigan EA, Fortinsky RH. Interrater reliability of the outcomes and assessment information set: results from the field. Gerontologist. 2004;44(5):689-692.

35. Readmissions Reduction Program (HRRP). CMS website. https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program.html. Updated April 18, 2016. Accessed February 3, 2017.

36. Long KH, Moriarty JP, Mittelman MS, Foldes SS. Estimating the potential cost savings from the New York University Caregiver Intervention in Minnesota. Health Aff (Millwood). 2014;33(4):596-604. doi: 10.1377/hlthaff.2013.1257.

37. Klug MG, Halaas GW, Peterson ML. North Dakota assistance program for dementia caregivers lowered utilization, produced savings, and increased empowerment. Health Aff (Millwood). 2014;33(4):605-612. doi: 10.1377/hlthaff.2013.1061.