Brian Chen, JD, PhD

To assess whether a care coordination and caregiver support intervention reduced use of acute medical services for both patients with Alzheimer disease (AD) and their caregivers.

Data were collected from patients with AD (n = 101) and their caregivers (n = 63) at Greenville Health System (now Prisma Health) in late 2012. Their data were linked to secondary all-payer claims data in South Carolina between 2011 and 2014.

 We conducted both a difference-in-differences regression and segmented regression analysis on the patients’ health care utilization patterns pre- and post intervention. Propensity score matching identified a control group composed of nonintervention patients with AD in South Carolina (n = 928). We examined caregiver differences via 

 Overall, the Memory Program did not reduce acute medical services. However, program participants experienced increases in total charges ($5243; 95% CI, $977-$9510) and in inpatient admissions with AD as a diagnosis (0.15; 95% CI, 0.029-0.272) but no increase in total all-cause charges. Intervention patients also had fewer emergency department (ED) visits (–0.0538; 95% CI, –0.102 to –0.0052) in some analyses. Finally, results suggest that post intervention, caregivers had half as many acute visits with depression as a diagnosis (from 0.22 to 0.11, difference of 0.11; 95% CI, –0.242 to 0.0198).

Although care coordination did not decrease overall acute health services use, coordination improved clinical documentation of patients’ memory impairment. ED visits may have begun to decrease among patients. Finally, stress levels may have fallen among caregivers.

Care coordination and caregiver support remain the primary intervention to meet the growing challenge of caring for patients with Alzheimer disease and related dementias (ADRD). Few studies, however, assessed their impact on objective measures of health care utilization. We studied patients and caregivers enrolled in the Memory Program in Greenville, South Carolina, and found strong evidence that the program led to better documentation of patients’ AD diagnosis. We also found evidence suggestive of a reduction in emergency department (ED) utilization among patients with AD and a potential reduction in urgent medical utilization for depression among caregivers.

Existing literature on ADRD interventions often focused on feasibility and self-reported outcomes.

Our studies assessed the impact of the Memory Program on objective measures of health care utilization for patients with AD.

The immediate impact may be better clinical documentation of AD even when patients seek care for other medical conditions.

There is suggestive evidence that the intervention reduced ED utilization among patients and acute medical service use for depressive symptoms among caregivers.

Alzheimer disease and related dementias (ADRD), or degenerative diseases of the brain, are collectively the sixth leading cause of mortality in the United States.

 These diseases exact a pronounced economic burden on society. In 2019, 5.8 million individuals lived with ADRD in the United States, with ADRD-attributed care costs exceeding $290 billion and projected to increase to $1.1 trillion by 2050.

Caregivers of patients with ADRD also experience a significant financial and health burden. Approximately 16 million Americans provide uncompensated care valued at more than $234 billion.

 In addition, depression among ADRD care partners is as high as 50%,

 with mortality rates for elderly spousal care partners approximately 63% higher than average.

These challenges require care model innovations that address both caregiver support and care coordination.

 Although no consensus definition exists,

 the Agency for Healthcare Research and Quality defines care coordination as “…the deliberate organization of patient care activities between [2] or more participants (including the patient) involved in a patient’s care to facilitate the appropriate delivery of health care services.”

 Presently, patients and caregivers face fragmented care delivery, and caregivers are often overwhelmed by the dual burden of navigating a complex health care landscape and providing direct care to patients with ADRD. Several studies over the past decade that evaluated care coordination programs provided strong evidence for feasibility and acceptability,

 caregiver satisfaction and improved quality of life,

 and delayed placement of patients with ADRD into nursing home facilities.

 However, the literature is mixed with respect to the relationship between care coordination and acute medical service use. One study reported a positive effect on reduced health care utilization,

 Moreover, many of the existing studies focused on care coordination–only models, without caregiver support.

 Thus, it remains unclear whether care coordination and care support in concert can reduce acute medical services use among patients with ADRD. Furthermore, we know of no study assessing the impact of a cognitive intervention program on health care utilization and costs among caregivers.

Our study addresses this knowledge gap by evaluating the impact of a continuum of care and caregiver support model on the health care expenditures and utilization patterns of both patients with AD and their caregivers. We leveraged data generated by a Memory Program implemented in 2012 at Prisma Health (then known as Greenville Health System) in Greenville, South Carolina. Modeled after the MemoryCare program in Asheville, North Carolina,

 the Memory Program prioritizes meeting both caregiver and patient needs by assisting with communication with the patients’ primary care physicians; providing licensed social worker services and other health care professionals for support, coaching, education, and referrals; creating a Caregiver Resource Library, an information source to learn about caregiving, managing stress, giving care, and topics related to memory health and self-care; and disseminating educational and supportive events led by experts in memory and geriatric care. This intervention afforded us an opportunity to examine the impact of a dual patient-caregiver support program on health care utilization and spending.

The study population consists of 101 patients with Alzheimer disease (AD) and 63 caregivers admitted to the 2012 Memory Program pilot program. Using a probabilistic matching algorithm, the South Carolina Revenues and Fiscal Affairs office (SC RFA) linked Memory Program patients and caregivers to all of their South Carolina emergency department (ED) and inpatient care claims between 2011 and 2014. Additionally, we obtained data on all nonintervention patients with AD (any individual with an 

International Classification of Diseases, Ninth Revision

] code of 331.0 in South Carolina other than patients at Prisma Health) between 2011 and 2014.

SC RFA deidentified the analytical sample and assigned a unique random identifier number to each patient and caregiver to track subjects over time. The deidentified data were then released to the research team for analysis.

We evaluated the impact of the Memory Program on 4 outcome variables of interest: total medical expenditures (charges), ED encounters, inpatient admissions, and discharge to a skilled nursing facility (SNF), aggregated at the patient-quarter level. We also created the same 4 outcome variables for AD-specific medical utilization by identifying only those visits associated with an 

 diagnosis code of 331.0. Finally, solely for caregivers, we used whether they had acute medical services for 6 comorbidities (diabetes, heart disease, cancer, depression, hypertension, and psychoses) as dependent variables to assess the impact of the Memory Program on caregiver health.

Propensity score matching was used to select non-Prisma patients with AD observably similar to patients in the Memory Program. Specifically, we matched patients on baseline (2012) age, gender, race, rural residence, type of insurance coverage, and 5 comorbidity indicators (diabetes, cancer, hypertension, depression, and congestive heart failure), as well as numbers of ED, outpatient, and inpatient visits. We used nearest neighbor matching, selecting 5 patients to serve as control for each patient with AD admitted to the Memory Program. Caregivers do not have a control group, as it is not possible to identify persons serving as caregivers from a medical claims database.

We employed the following difference-in-differences analyses to test the impact of the Memory Program on AD health care utilization and charges, using propensity score–matched non–Memory Program patients with AD as controls.

 represents the various outcomes of interest for individual 

 identifies whether the observation is in the postintervention period (2013 and 2014), and 

 indicates participation in the Memory Program. The key coefficient is 

), which represents the impact of the Memory Program relative to the control group. We chose to include patient-level fixed effects (

) to control for unobserved time-invariant characteristics. Doing so causes the treatment variable to drop out of the estimation because it is time-invariant. However, we are still able to estimate the primary coefficient of interest, or 

In addition, we also used a segmented regression model for patients. This procedure compared the outcome variables for the subjects before and after joining the Memory Program, as follows:

Here, all the variables are identical to the difference-in-differences, except the following: The 

simply counts the number of quarters that have elapsed since the beginning of the study period. The 

 remains 0 until Memory Program enrollment, and then counts the number of quarters that have passed after enrollment. The main coefficients of interest are 

, which, respectively, represent any discontinuity in the outcome variable at enrollment and postperiod changes in outcome trends relative to the preperiod trends.

Owing to the lower frequency of acute events among caregivers, a segmented regression model would likely estimate inaccurate trends using highly stochastic data. We therefore conducted simple 

 tests comparing the pre– and post–Memory Program outcomes for caregivers.

Our results reveal 3 primary patterns. First, Memory Program patients had a significant increase in AD-related charges and inpatient admissions compared with controls, but there was no increase in all-cause charges and inpatient admissions. Second, some analyses show that Memory Program patients may have had a decrease in all-cause and AD-related ED visits. Finally, caregivers did not experience a reduction in serious health outcomes, as measured by inpatient and ED utilization. However, caregivers may have experienced a reduction in depression-related acute services.

Patients’ baseline characteristics suggest that Prisma Memory Program patients differed from the general population with AD in South Carolina (

). Memory Program patients were older (76.67 vs 64.02 years), less likely to be Black (21% vs 25%), less likely to have rural residence (2% vs 24%), more likely to have Medicare (81% vs 63%), and less likely to have Medicaid (3% vs 9%). At baseline, Prisma patients had slightly lower average total health care charges ($58,286.08 vs $65,549.85) and similar health care utilization but fewer inpatient admissions (0.95 vs 1.24). Prisma patients were less likely to have diabetes (21% vs 25%) but more likely to have hypertension (78% vs 61%).

Caregivers were, on average, 15 years younger than care recipients (61.83 vs 76.67 years) and were more likely to have commercial health insurance (46%) than Medicare (37%). Not surprisingly, caregivers were younger and healthier than patients, although it is noteworthy that they had the same rate of depression as patients.

Table 1 (see “Matched Controls”) provides the results of matching patients based on patient and clinical characteristics. Matched patients have better covariate balance, particularly for age, rural residence, number of inpatient admissions at baseline, Medicare coverage, and commercial insurance. (See the 

] for additional analyses on bias reduction through propensity score matching.) Overall, our nearest-neighbor propensity score matching reduced the likelihood ratio χ

 value from 157.7 to 2.23 and the mean bias from 22.4 to 5.1.

Difference-in-Differences Results for Patients With AD

 (panel A) describes our difference-in-differences results using our matched patients with AD as controls. Enrollment in the Prisma Memory Program was associated with little or no change in all-cause charges, inpatient and ED utilization, or the likelihood of posthospitalization discharge to an SNF. However, Prisma patients were more likely to have their charges and inpatient admissions ascribed to an AD-related cause than control patients. Prisma patients had an average of $5243 (95% CI, $77-$9510) more in AD-related charges and 0.15 (95% CI, 0.0293-0.272) more inpatient admissions per quarter than non-Prisma controls with AD.

The difference-in-differences results are subject to bias if the control group is not properly selected. We conducted 2 alternative analyses to assess the robustness of our results. First, we simply ran an additional difference-in-differences regression using all non-Prisma patients with AD as controls. The results are broadly similar to those of the analysis using propensity-score matched controls (see Table 2, panel B). Second, we used segmented regression analyses to estimate changes in outcomes around the enrollment date using only the intervention group without using a comparison group. Analyses still reveal a statistically significant increase in total AD-related charges after Memory Program enrollment ($4310; 95% CI, $2-$8618). There was also an increase in AD-related inpatient admissions of 0.130, which is similar in magnitude to other specifications, but the result is not statistically significant. However, our segmented regression analysis shows that AD-related ED utilization decreased by 0.0538 (95% CI, –0.102 to –0.00521) visits per quarter compared with preenrollment trends. Because this coefficient represents the change in trends, the actual slope of the trend in postintervention AD-related ED visits is a 

 of 0.0285 visits (ie, 0.0253 [preintervention slope] + –0.0538 [change in postintervention slope] = –0.0285; 95% CI, –0.0564 to –0.00727) (see Table 2, panel C).

Test of Differences in Means for Caregivers

For caregivers, the results show no statistically significant difference in means between the pre– and post–Memory Program measures for total charges, ED visits, and inpatient admissions. 

 tests on differences in means for comorbidities revealed no statistically significant comparisons at α < 0.05. However, the proportion of caregivers using acute medical services for depression decreased from 22% to 11% (95% CI, –0.242 to 0.0198; 

Our analyses yielded 3 principal findings. First, Memory Program patients were more likely to have their AD diagnosis documented during acute medical visits. Overall ED and inpatient admissions did not increase among Memory Program patients—only the likelihood of recording AD as part of the patient’s present-at-admission diagnoses increased. This superior documentation may have resulted from improved care coordination, more caregiver communication, or a combination of these factors. Although seemingly inconsequential, comorbidity documentation improvement—a care coordination indicator—is an essential antecedent to improved health outcomes.

Theoretically, better documentation may enhance chronic disease management by reducing missing clinical information, improving diagnostic accuracy, and facilitating informational exchange necessary for efficient care coordination.

 Accurate and centralized health information reduces medication errors

 and informs root cause analysis of adverse events.

 Documentation and coding accuracy are especially critical for complex patients with multimorbidity who require coordinated care plans involving multiple clinical stakeholders and departments. Empirical data suggest that patients with ADRD have the highest levels of multimorbidity among any patient group with chronic conditions.

 This observation is not accidental, as age

 is a risk factor for multiple comorbidities,

 and both age and multimorbidity are risk factors for dementia.

 Superior documentation may therefore be especially important for the care of patients with ADRD.

Second, our analyses revealed no differences overall in acute health care utilization between Memory Program and control patients. This finding is consistent with those of earlier meta-analyses, including a study by Pimouguet et al

 finding no evidence that care coordination reduced health care expenditures or hospitalizations. However, like ours, prior studies also found evidence suggestive of better documentation of ADRD. The MIND trial (a randomized controlled trial of an 18-month care coordination program for ADRD) reported no increase in total outpatient services but detected an increase in ADRD-related outpatient visits.

 The concordance of our results suggests that care coordination interventions may very well improve clinical documentation of AD patients’ health care utilization, even without reducing immediate health care utilization and costs.

Nevertheless, in the segmented regression analyses, we observed that all-cause and AD-specific ED visits began declining among Memory Program patients toward the end of the study period. Boustani et al similarly observed less frequent utilization of acute health services.

 However, a comprehensive literature search revealed that these results are outliers and that few published studies reported reduced ED utilization. More commonly, studies documented that care coordination for AD satisfied unmet needs, increased quality of life, and improved quality of care without necessarily reducing health care utilization or costs.

Future research should include extended follow-up to confirm or qualify our findings, but there are at least 2 reasons for the general lack of a finding that care coordination produces theorized benefits. First, care coordination is costly to administer and addresses both overutilization and underutilization resulting from fragmented care.

 As such, care coordination can actually increase health care services utilization when previously unmet needs are addressed with better clinical coordination. Second, there is no consensus definition of care coordination, so it is likely inconsistently applied in clinical settings.

 In breaking down care coordination into coordination activities (eg, communication and monitoring) and interdependencies (eg, flow, shared resources, simultaneity), Kianfar et al identified at least 258 care coordination activities.

 Petersen et al, in a meta-analysis, found at least 37 widely diverse theoretical frameworks for care coordination, likely with different theory-to-practice links.

 With such a broad range of potential implementations of care coordination, it is possible that existing studies have evaluated multiple variations of clinical care coordination models, most of which have not yielded reductions in health care utilization and costs.

Third, our analyses did not find that caregivers experienced a statistically significant decrease in acute health services use. However, our analyses suggest that caregivers experienced reduced ED use and inpatient care for depression. This result shows that 22% of caregivers experienced an acute episode of depression requiring urgent care in the preintervention period, whereas only 11% had such an incident in the postintervention period. The results are significant only at the .10 level, and the limited number of caregivers in the preintervention period (63) potentially failed to provide sufficient power to detect a significant result at the .05 level.

These results are consistent with those of multiple published studies reporting caregiver satisfaction, stress reductions, and improvements in quality of life. For example, the investigators of the MIND trial observed that the intervention reduced the time that caregivers were required to spend with care recipients.

 The Veterans Health Administration’s Partners in Dementia Care program showed that intervention group caregivers experienced significant improvements in meeting previously unmet needs, alleviating 3 types of caregiver strains, as well as decreasing depression and increasing access to caregiver support services.

 Similarly, in the Veterans Health Administration’s Resources for Enhancing Alzheimer’s Caregiver Health randomized controlled trial, caregivers reported decreased burden, depression, and caregiving frustrations, all of which affected quality of life.

 Other trials also observed that caregiver support reduced unmet needs,

 increased caregiver self-efficacy, and reduced caregiver burden.

 Our results affirm the findings of these prior studies and contribute to the literature by using objective health services measures in claims data, beyond caregiver self-reports.

We acknowledge several study limitations. First, our study is not a randomized controlled trial. Existing patients with AD and their caregivers opted to enroll in the Memory Program for a fee. Our study group in fact consisted of patients with AD who were more likely to have advanced illness than the general patient population with AD, but we attempted to alleviate this selection bias by using propensity score matching to identify non-Prisma patients with AD in South Carolina observably most similar to our intervention group. Second, data deidentification resulted in losing/obscuring the exact enrollment dates for each individual AD patient and caregiver. We chose to sacrifice some precision to preserve participant protection principles, noting that we were able to identify a quarter in which a substantial majority of our subjects enrolled in the Memory Program. We designated this quarter as the intervention start period. Potential error introduced by a misassigned start date could introduce noise and reduce statistical significance. Despite this, our analyses detected statistically significant relationships in several specifications.

ADRD incidence and prevalence continue to increase worldwide, with significant health and economic implications for patients with ADRD, their caregivers, health systems, and policy makers. Absent a cure, optimal care coordination and caregiver support remain the prime media to improve ADRD patient and caregiver welfare. Existing literature generally considers only program feasibility and self-reported outcomes of behavior symptoms, caregiving support, and caregiver quality of life. Fewer studies have assessed program impact on health care utilization. Our study showed that Memory Program enrollment was associated with better clinical documentation of patients’ AD diagnosis. Moreover, we found suggestive evidence that the program reduced ED utilization among patients with AD and possibly reduced urgent medical utilization for depression among caregivers. Future studies on the impact of care coordination and caregiver support programs should consider extended follow-up periods to assess the long-term benefits from these care delivery interventions.


 Department of Health Services Policy and Management, Arnold School of Public Health, University of South Carolina (BC, XC, MFH, DJ, NH), Columbia, SC; School of Psychology, Northcentral University (BS-L), San Diego, CA; Prisma Health Midlands (MFH), Greenville, SC.

 This publication resulted from research supported by the Institute for the Advance of Health Care at the Greenville Health Systems (now Prisma Health). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agency.

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (BC, BS-L); acquisition of data (BC, BS-L); analysis and interpretation of data (BC, XC, NH); drafting of the manuscript (BC, BS-L, MFH, DJ); critical revision of the manuscript for important intellectual content (BC, XC, MFH, DJ, NH); statistical analysis (BC, XC, NH); provision of patients or study materials (BS-L); obtaining funding (BC, BS-L); administrative, technical, or logistic support (BS-L, MFH, DJ); and supervision (BC).

 Brian Chen, JD, PhD, Department of Health Services Policy and Management, Arnold School of Public Health, University of South Carolina, 915 Greene St, Ste 354, Columbia, SC 29208. Email: bchen@mailbox.sc.edu.

1. Alzheimer’s disease fact sheet. National Institute on Aging. May 22, 2019. Accessed February 14, 2020. https://www.nia.nih.gov/health/alzheimers-disease-fact-sheet

2. Facts and figures. Alzheimer’s Association. Accessed February 14, 2020. https://www.alz.org/alzheimers-dementia/facts-figures

3. 2019 Alzheimer’s disease facts and figures. Alzheimer’s Association. 2019. Accessed February 14, 2020. https://www.alz.org/media/documents/alzheimers-facts-and-figures-2019-r.pdf

4. Joling KJ, van Hout HPJ, Schellevis FG, et al. Incidence of depression and anxiety in the spouses of patients with dementia: a naturalistic cohort study of recorded morbidity with a 6-year follow-up. 

. 2010;18(2):146-153. doi:10.1097/JGP.0b013e3181bf9f0f

5. Gainey RR, Payne BK. Caregiver burden, elder abuse and Alzheimer’s disease: testing the relationship. 

6. Dening KH, Aldridge Z, Pepper A, Hodgkison C. Admiral Nursing: case management for families affected by dementia. 

. 2017;31(24):42-50. doi:10.7748/ns.2017.e10600

7. Backhouse A, Dickens C, Richards D, McCabe R. Key components in models of community-based interventions coordinating care in dementia: a mixed studies systematic review protocol. 

. 2015;4:156. doi:10.1186/s13643-015-0143-y

8. Galvin JE, Valois L, Zweig Y. Collaborative transdisciplinary team approach for dementia care. 

. 2014;4(6):455-469. doi:10.2217/nmt.14.47

9. Jamieson M, Grealish L, Brown JA, Draper B. Carers: the navigators of the maze of care for people with dementia—a qualitative study. 

. 2014;15(5):1112-1123. doi:10.1177/1471301214554930

10. Schultz EM, McDonald KM. What is care coordination? 

. 2014;17(1-2):5-24. doi:10.1177/2053435414540615

11. McDonald KM, Sundaram V, Bravada DM, et al. 

Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 7: Care Coordination)

. Agency for Healthcare Research and Quality; 2007.

12. Callahan CM, Boustani MA, Weiner M, et al. Implementing dementia care models in primary care settings: the Aging Brain Care Medical Home. 

. 2011;15(1):5-12. doi:10.1080/13607861003801052

13. Judge KS, Bass DM, Snow AL, et al. Partners in Dementia Care: a care coordination intervention for individuals with dementia and their family caregivers. 

. 2011;51(2):261-272. doi:10.1093/geront/gnq097

14. D’Souza MF, Davagnino J, Hastings SN, Sloane R, Kamholz B, Twersky J. Preliminary data from the Caring for Older Adults and Caregivers at Home (COACH) program: a care coordination program for home-based dementia care and caregiver support in a Veterans Affairs Medical Center. 

. 2015;63(6):1203-1208. doi:10.1111/jgs.13448

15. Samus QM, Black BS, Reuland M, et al. MIND at Home–Streamlined: study protocol for a randomized trial of home-based care coordination for persons with dementia and their caregivers. 

. 2018;71:103-112. doi:10.1016/j.cct.2018.05.009

16. Reuben DB, Evertson LC, Wenger NS, et al. The University of California at Los Angeles Alzheimer’s and Dementia Care program for comprehensive, coordinated, patient-centered care: preliminary data. 

. 2013;61(12):2214-2218. doi:10.1111/jgs.12562

17. Dang S, Remon N, Harris J, et al. Care coordination assisted by technology for multiethnic caregivers of persons with dementia: a pilot clinical demonstration project on caregiver burden and depression. 

. 2008;14(8):443-447. doi:10.1258/jtt.2008.080608

18. Engelhardt JB, Kisiel T, Nicholson J, Mulichak L, DeMatteis J, Tobin DR. Impact of a care coordination and support strategic partnership on clinical outcomes. 

. 2008;26(3):166-172. doi:10.1097/01.NHH.0000313348.85980.1e

19. Clevenger CK, Cellar J, Kovaleva M, Medders L, Hepburn K. Integrated Memory Care Clinic: design, implementation, and initial results. 

. 2018;66(12):2401-2407. doi:10.1111/jgs.15528

20. Tanner JA, Black BS, Johnston D, et al. A randomized controlled trial of a community-based dementia care coordination intervention: effects of MIND at Home on caregiver outcomes. 

. 2015;23(4):391-402. doi:10.1016/j.jagp.2014.08.002

21. Bass DM, Judge KS, Snow AL, et al. Caregiver outcomes of Partners in Dementia Care: effect of a care coordination program for veterans with dementia and their family members and friends. 

. 2013;61(8):1377-1386. doi:10.1111/jgs.12362

22. Vickrey BG, Mittman BS, Connor KI, et al. The effect of a disease management intervention on quality and outcomes of dementia care: a randomized, controlled trial. 

. 2006;145(10):713-726. doi:10.7326/0003-4819-145-10-200611210-00004

23. Nichols LO, Chang C, Lummus A, et al; Resources for Enhancing Alzheimer’s Caregivers Health II Investigators. The cost-effectiveness of a behavior intervention with caregivers of patients with Alzheimer’s disease. 

. 2008;56(3):413-420. doi:10.1111/j.1532-5415.2007.01569.x

24. Nichols LO, Martindale-Adams J, Burns R, Graney MJ, Zuber J. Translation of a dementia caregiver support program in a health care system—REACH VA. 

. 2011;171(4):353-359. doi:10.1001/archinternmed.2010.548

25. Fortinsky RH, Unson CG, Garcia RI. Helping family caregivers by linking primary care physicians with community-based dementia care services: the Alzheimer’s Service Coordination Program. 

. 2002;1(2):227-240. doi:10.1177/147130120200100208

26. Meeuwsen EJ, Melis RJF, Van Der Aa GCHM, et al. Effectiveness of dementia follow-up care by memory clinics or general practitioners: randomised controlled trial. 

27. Pimouguet C, Lavaud T, Dartigues JF, Helmer C. Dementia case management effectiveness on health care costs and resource utilization: a systematic review of randomized controlled trials. 

. 2010;14(8):669-676. doi:10.1007/s12603-010-0314-4

28. Koch T, Iliffe S, Manthorpe J, et al; CARE-DEM. The potential of case management for people with dementia: a commentary. 

. 2012;27(12):1305-1314. doi:10.1002/gps.3783

29. Tam-Tham H, Cepoiu-Martin M, Ronksley PE, Maxwell CJ, Hemmelgarn BR. Dementia case management and risk of long-term care placement: a systematic review and meta-analysis. 

. 2013;28(9):889-902. doi:10.1002/gps.3906

30. Boustani MA, Sachs GA, Alder CA, et al. Implementing innovative models of dementia care: the Healthy Aging Brain Center. 

. 2011;15(1):13-22. doi:10.1080/13607863.2010.496445

31. Lee L, Hillier LM, Harvey D. Integrating community services into primary care: improving the quality of dementia care. 

32. MemoryCare. Accessed November 3, 2019. https://memorycare.org

33. Hahn KA, Ohman-Strickland PA, Cohen DJ, et al. Electronic medical records are not associated with improved documentation in community primary care practices. 

. 2011;26(4):272-277. doi:10.1177/1062860610392365

Articles

The authors investigated whether patient coordination and caregiver support for Alzheimer disease reduced health care utilization and expenditures among enrollees in the Memory Program in South Carolina.

Effect of Care Coordination on Patients With Alzheimer Disease and Their Caregivers

On March 1, 2016, the Supreme Court issued a 6-2 decision in 

, holding that the federal Employee Retirement Income Security Act (ERISA)—a comprehensive federal law that regulates a variety of employee benefits, including health plans established by employers&mdash;preempts a Vermont law requiring self-funded employer health insurers to submit certain types of health data to state databases.

 Vermont had passed the law to collect and aggregate state-level health data, with the goal of understanding and addressing rising healthcare costs.

 As increasing numbers of employees receive coverage from self-funded employers, their healthcare data have become important to ensure transparency and to craft state-level healthcare policy. Although some laud the Supreme Court decision as a win for medical privacy, the case highlights the challenges in managing conflicting state and federal laws, and balancing patient privacy interests against the state’s interests in healthcare reform.

Background: Vermont’s Law, its Opposition, and the Court’s Holding

Enacted in 2005, the Vermont law aimed to improve healthcare quality by collecting various health-related data, including “medical claims data, pharmacy claims data, member eligibility data, provider data and other information.”

 In particular, the law required self-funded insurance providers to submit, upon request, healthcare information on prices and annual registration forms, and to report claims at specified intervals.

 In the recent past, a growing number of states have established databases that collect health insurance claims information from all healthcare payers into a statewide information repository. By January 2016, Vermont and 17 other states had enacted “all-payer claims databases,” while more than a dozen others considered such a law or program.

Liberty Mutual Insurance Company, which administers self-insured health plans for over 80,000 individuals nationwide,

 objected to the Vermont law, claiming its disclosure requirements violate its fiduciary duties under ERISA.

 Vermont had issued a subpoena to demand that Liberty Mutual release medical and pharmacy claim files pursuant to Vermont law, and the company responded by filing suit—claiming that ERISA preempted Vermont’s data-reporting requirement for self-insured plans. The district court sided with Vermont, but the Second Circuit Court of Appeals reversed the lower court’s decision and ruled in favor of Liberty Mutual.

 The Second Circuit placed a heavy emphasis on the privacy risks associated with this type of reporting requirement and found that such privacy risks would burden beneficiaries of the plans.

The Supreme Court subsequently granted certiorari, and, with little difficulty, found that ERISA did preempt the Vermont law.

 The Court noted that the only relevant question is whether Congress intended ERISA to preempt state legislation, such as Vermont’s. Because the Vermont law both “govern[ed]…a central matter of plan administration” and “interfere[ed] with nationally uniform plan administration,” the Court found that the law was of the very type that Congress intended ERISA to preempt.

 To allow states to have their own reporting requirements would “impos[e] novel, inconsistent, and burdensome reporting requirements on plans.”

Unaddressed Controversy: Should Privacy or Data Drive Healthcare Policy?

Unlike the Second Circuit’s ruling, the Supreme Court’s opinions (majority, concurring, and dissenting) made little of the privacy aspect, noting only that Liberty Mutual challenged the law to protect the privacy of individual medical records.

 None of the justices, however, expressed any issue with government collection of individual healthcare information.

 The majority opinion even suggested that the Secretary of Labor could require ERISA plans to report individual claims data to a national or state database, indicating that the Court saw no privacy issues at stake, so long as statutes and policies meet the requirements of ERISA.

Although the Court barely addressed the issue, many have recognized an inherent conflict between protecting medical privacy and collecting healthcare information to shape effective policy. Eighteen states have created such databases to determine the nature and distribution of healthcare costs, in hopes of implementing policies and regulations that could reduce costs.

 As Justice Ruth Bader Ginsberg noted in her dissent in 

, “Stopping states from collecting claims data from self-insured employer healthcare plans would thus hugely undermine the reporting regimes on which Vermont and other States depend to maintain and improve the quality, and hold down the cost, of healthcare services.”

, states can no longer mandate data reporting from the majority of employees nationwide, given that about 63% of covered workers are enrolled in a plan that is partially or entirely self-funded.

Likewise, the American Hospital Association noted that it is essential that self-funded insurers be included in all-payer databases if those databases are to realize their potential.

 The American Hospital Association and the Association of American Medical Colleges agreed that hospitals have only the data for the patients they treat, but information from an individual’s broad healthcare experience is needed to inform clinical, payment, and public health policy.

 Complete data is necessary to conduct useful analyses of the healthcare system, across its entire spectrum, to make meaningful changes. Striking down these laws may also reduce the transparency necessary to encourage insurers participating in the healthcare marketplace and improve overall state health.

Opponents of the law, however, fear potential data breaches in the large-scale transmission and storage of personal health data, and suggest that any savings and increased efficiency will come at a high cost. Thus, said Twila Brase, president of Citizens’ Council for Health Freedom, in reaction to the ruling, “Today is a good day for medical privacy in America.”

 The ruling means that fewer patient medical records are at risk of privacy violation.

 Many, including the Association of American Physicians and Surgeons (AAPS) believe that patient privacy cannot adequately be protected in large government databases.

 Large, centralized databases are vulnerable to hacking, and are targets for cyber-attacks, putting the private medical information of millions of Americans at risk of public disclosure.

The Second Circuit explicitly acknowledged that the laws “require[…] ERISA plans to record, in specified format, massive amounts of claims information and to report that information to third parties, creating significant (and obvious) privacy risks and financial burdens.”

 Moreover, de-identification may do little to protect individual privacy; researchers at Harvard demonstrated that de-identified private medical information can be re-identified.

 The potential public disclosure of this private information, through breach or re-identification, could adversely impact careers and family lives, causing a disincentive against seeking care for stigmatized medical conditions. The AAPS believes that ultimately, the medical profession, rather than “big brother” states, is the appropriate actor to spearhead healthcare cost reductions.

Implications for the Future of Healthcare Data Collection After 

 hampers efforts to track the quality and cost of care, it implies that, at least for now, workers in America covered under self-insured plans are protected from the mandatory release of their private medical records to state governments. Advocates of privacy rights are not alone in celebrating the decision.

 Many employers and health insurers see the decision as a significant victory for self-funded insurers who otherwise would have to comply with conflicting patchworks of state laws and regulations.

 Without the ruling, large self-funded employers that operate in multiple states could face a myriad of different regulations in each state, leading to high administrative costs and conflicts in compliance requirements. The new ruling alleviates that burden by ensuring that self-insured employers all face one uniform set of national regulations. Uniformity in reporting helps with efficient plan administration.

The majority opinion did, however, suggest a way in which states might obtain the information they desire—through the federal government.

 The Secretary of Labor could, for example, promulgate regulations to require self-funded insurers to report healthcare data to the federal government, who could then distribute this information to the states.

 Moreover, the ruling does not prevent self-funded insurers from voluntarily reporting health information to the states.

These technical workarounds to comply with the Court’s decision, however, may overlook the risk of privacy violation. Although many can appreciate the need for centralized healthcare databases and the critical analysis regarding the costs and quality that they enable for policy formulation, the Court’s reticence on privacy issues associated with data collection may suggest that it finds no inherent privacy violation in massive, government-collected databases of private medical information.

 was silent on the privacy issues surrounding health databases and the ruling was merely a technical application of the doctrine of federal preemption, according to which federal law trumps state law when the two are in conflict, this case may pave the way for the federal government to collect health plan data or to approve or delegate to a state the authority to collect such data. As one physician put it, however, “[W]e need to foster a public debate over citizens’ right to medical privacy versus government seizure of private information without consent, because any loss of privacy rights undermines our society, and threatens our freedoms and our way of life.”

 A rigorous analysis of the risks and benefits of government-mandated healthcare databases, and investment in improved technology to protect such data already held by the government, are of utmost importance.

 Department of Health Administration and Policy, George Mason University (YTY), Fairfax, VA; Department of Health Services Policy and Management, University of South Carolina (BC), Columbia, SC. 

Gobeille v. Liberty Mutual Insurance Company highlights the challenges in managing conflicting state and federal laws, and balancing patient privacy interests against the state’s interests in healthcare reform.

Brian Chen, JD, PhD

Articles

Effect of Care Coordination on Patients With Alzheimer Disease and Their Caregivers

Medical Record Privacy and State Health Reform After Gobeille v. Liberty Mutual Insurance Co.