Home-Based Primary Care: Can We Take These Results to the Bank?

To some, “home-based primary care” (HBPC) may evoke images of frail shut ins — unable to leave their homes due to physical, mental, or cognitive impairments.  Others imagine a nostalgic past when physicians made house calls to acutely ill patients. The aging of America is no secret.¹ With aging, comes an increased burden of disease and pharmacotherapy.² Those with multiple chronic conditions are high utilizers of health care resources.^3-5  A small number of Medicare beneficiaries incur a disproportionately large fraction of total healthcare expenditures.⁵ Demonstration projects in home-based primary care (HBPC) are currently underway as part of the Accountable Care Act’s Independence at Home (IAH) program,⁶ with the goal of decreasing overall costs.

There has been considerable enthusiasm for HBPC as a model of care to rectify our fragmented healthcare system. But does HBPC make a difference for either patients or payers?  HBPC-based trials examining their impact on hospital admissions^7-12, emergency department visits^7,13,14, long-term care/skilled-nursing facility placement^10,11,14 and quality of life/patient satisfaction¹³ have all failed to show a positive impact on patient-oriented outcomes.  And, from the payer’s perspective, the impact of HBPC on the cost of care has varied widely.^8,11,13

Most recently, De Jong and colleagues¹⁵ conducted a case-control cohort study that examined the impact of a HBPC service on costs and survival in a fee-for-service Medicare population. Cases were individuals over the age of 65, newly enrolled in the MedStar Washington Hospital Center HBPC system between 2004 and 2008,          and who did not have HMO coverage at baseline. Excluding those with HMO coverage gave a clearer picture of individuals who solely used Medicare Parts A and B for their health-care coverage, as well as provided complete medical datasets from a sole payer. Medicare Part D information was not included. Individuals were excluded if they resided in a LTCF, but could be included if temporarily admitted to a LTCF for skilled nursing care. Cases were matched to controls in a 1:3 ratio based on multiple variables, including age and comorbidities, from a pool of beneficiaries from surrounding metro areas. Univariate analysis was utilized for the primary outcomes of total Medicare Part A & B costs, use of medical services, and mortality. Total costs were calculated based on the following categories: hospice, home health, physician and non-physician practitioner billing, skilled nursing care, hospitalizations, diagnostic tests, transportation, Medicare Part B drugs, durable medical equipment, and outpatient facility use. A multivariable regression was performed to account for baseline differences including age, frailty, and selected major chronic diseases.

The HBPC team consisted primarily of physicians, some of whom were geriatricians, nurse practitioners, social workers, nurses, and home health aides. Home visits and services were performed based on the level of care deemed necessary for each patient.  A pharmacist and mental health professional joined the team for weekly team meetings.  Patients were followed by the team during inpatient stays and care transitions. An integrated electronic health record (EHR) linking inpatient and outpatient data was utilized as well as home-based diagnostic tools. Patients were followed until the end of the study period (2013), were placed in a LTCF, entered a Medicare Advantage (HMO) plan, or passed away.

The majority of the participants were female (76.7%) and African American (90.2%).  Cases were slightly older than the controls (83.7 vs. 82.0 years; p < 0.001). In the 4-month period prior to HBPC enrollment, there was no significant difference in total Medicare spending but there were differences in some cost categories.

During an average follow-up of approximately 2 years from enrollment, total Medicare costs were lower for cases than controls ($44,455 vs. $50,977; p = 0.01). While certain expenses were higher for cases (home health care and hospice), cases had significantly fewer hospitalizations (P=0.001), ER visits (P=0.001), and SNF days (P=0.001). Cases had roughly twice the number of generalist practitioner visits during the study (P < 0.001). Multivariate analysis showed that total Medicare costs were reduced by 17% for those participating in the HBPC program versus matched controls, translating to a potential cost savings greater than $6 million over 2 years. A subgroup analysis found that the frailest patients were the primary driver of cost savings across the entire cohort.  Mortality rates did not differ between the groups or the mean time to death.  Overall, the rates of death were high in both groups.

In an era where hospitals and accountable care organizations must increasingly shoulder costs of care, these findings are intriguing. However, there are important limitations that impact the generalizability of these results. An integrated health system and EHR, allowing practitioners to follow patients both as inpatients and outpatients, is not the norm in the United States. While the patient population was older and female, it was also predominantly African American.  The health system included smaller, non-urban hospitals. Importantly, the investigators did not have access to Medicare Part D claims data.   Therefore, there is no information regarding the impact of this HBPC program on drug costs or how drug costs may have influenced the overall cost of care. Finally, control patients were not exclusively from the same geographic area.  Therefore differences in practice patterns and utilization may account for the findings.

Despite these limitations, this study offers some important insights. It is the first study attempting to compare outcomes most relevant to payers – Medicare in this case. The subgroup analysis, although only hypothesis generating, may be useful when developing algorithms to identify those who might benefit the most from HBPC.  The team’s involvement during care transitions is unique, as other HBPC interventions do not typically include coordination of care from one care venue to another.

There are several important considerations we should keep in mind for future studies:

• Study design: Prospective, randomized studies are needed to draw definitive conclusions regarding advantages and disadvantages of HBPC.
• Team composition: Standardization of HBPC inter-professional teams may have important implications for comprehensiveness of care, patient outcomes, and associated costs.
• Frequency of follow-up: Standardization in the frequency of HBPC visits, team meetings, and coordination of care transitions may be important contributors to outcomes such as quality of life and perceptions of access to care.
• Duration of follow-up: The long-term impact of HBPC on patient outcomes is poorly elucidated; the majority of studies have followed patients for less than one year after enrollment.  

Despite the positive impact of a HBPC intervention on cost in this study⁷, health systems may be reticent to embrace HBPC. This may be especially true in a fee for service Medicare plan where savings do not translate to tangible financial benefits to the health system providing the care – indeed, it might reduced revenue.  Employing an interprofessional team devoted solely to the care of home-based patients, as well as the costs of the technology to support such care, may be an insurmountable financial obstacle for some health systems.   And, if a health-system determined that HBPC was in their best interest, would they include a pharmacist on the interprofessional team?  It seems logical that a pharmacists would be the most appropriate individual to oversee medication use and identify medication-related problems.  But pharmacists do not yet have provider status and cannot bill Medicare.  While the data is far from clear, this analysis suggests that HBPC is a cost-effective approach to meeting the needs of the frailest in our society.   What do you think?