Use of the ICU in the United States, both overall and near the end of life, is rising (1–3) and has the potential to be a financial burden on patients and families. A recent survey conducted by “The Economist” and the Kaiser Family Foundation found that not burdening families with healthcare costs was patients’ highest-ranked priority for end-of-life care (4). A major rationale for health insurance in the United States has been protection against financial risk. In 1965, when President Johnson signed Medicare into legislation, he declared, “No longer will illness crush and destroy the savings that [older Americans] have so carefully put away over a lifetime.” One of the most robust findings of recent health insurance expansions is the decrease of medical bankruptcy and increase in financial security (5–8). However, spending on healthcare is increasingly being shifted to patients, through higher deductibles and increases in copayments. Between 2008 and 2014, out-of-pocket (OOP) health spending in the U.S. rose 11.5%, reaching $329.8 billion (9). This shifting of financial responsibility is happening within private insurance and within Medicaid, where states can impose copayments, coinsurance, and deductibles for some services for some beneficiaries (10).
Although one might think Medicare insulates individuals from these financial burdens, in international comparisons, the U.S. stands out for the presence of reported financial barriers to access to care (11). Further, Medicare’s rising burden on the national budget has led to numerous reform proposals, including changing Medicare to a premium support voucher program, which would lead seniors back into the private insurance market. This shift in the design of health insurance contracts decreases the extent to which insurance insulates individuals from the financial burden of healthcare.
Recent studies estimating the extent to which Medicare protects against the financial risk of illness have found conflicting results. Some studies found Medicare to be relatively insulating from catastrophic financial risk (12–14), whereas another concluded that elderly households face considerable financial risk in the last 5 years of life and that this risk varies by disease (15, 16). Little is known, however, about the extent of health-related financial risk that individuals who require intensive care near the end of life face, and how this risk varies by insurance status. This population is important because in the United States, approximately one in five persons die in or shortly after discharge from an ICU (17). Further, ICU use near the end of life is increasing (2).
This study contributes to the literature by estimating the OOP expenditures for direct medical care and health-related services that decedents who used the ICU near the end of life faced in the last 12 months of life, by type of care and by insurance coverage.
We used seven waves of post-death, or “exit,” interview data (2002–2014) from the Health and Retirement Study (HRS), a longitudinal study of aging that interviews approximately 20,000 adults over 50 every 2 years (18, 19). Exit interviews are conducted with a spouse, child, or other knowledgeable proxy for all HRS participants who died since the prior interview wave. We combined exit data with data from the last interview prior to death.
From the 9,230 decedents with exit interview data, we selected the 3,263 decedents who, per the exit interview, spent time in an ICU at some point between their last interview and death. Participants were excluded for unknown date of death (n = 5); no interview in the two waves prior to death (n = 148); delay in collecting the exit interview (n = 38); or missing covariate data (n = 163). Prior studies have found substantial to near perfect agreement between self-reported hospitalizations and routinely collected healthcare utilization data (20). To assess the validity of self-reported ICU use, we reviewed Medicare claims for eligible decedents who gave consent to link their medical records and were continuously enrolled in Medicare fee-for-service (FFS) for the entirety of their recall period (n = 1,735). Seventy-nine percent of those reporting an ICU stay had a claim indicating ICU use during the recall period.
During the exit interview, OOP costs were captured in the following categories: hospital, nursing home, hospice, home health, doctor visits, prescription drugs, and a special facility or service (e.g., outpatient rehabilitation). Expenses not captured in these categories were elicited in a final “other” OOP costs category (interview questions in eSupplement, Supplemental Digital Content 1, http://links.lww.com/CCM/E465). Due to the small number of respondents with OOP costs for hospice, home health, and a special facility or service, these categories were not included in our analyses.
For each category except for prescription drugs, proxies were asked about OOP costs incurred between the last interview and death for decedents that participated in the prior interview wave, or for the 2 years prior to death for decedents that had not completed the prior interview wave. In our sample, the exit interviews reflected a 15-month time period, on average, with a range of 1–34 months. (Table 1) OOP costs for prescription drugs reflected an average 1-month time period for all decedents. Total OOP costs were calculated by summing reported OOP costs in all categories; we multiplied prescription drug costs by the number of months in the recall period for comparability.
Depending on the cost category, a quarter to a third of respondents were unable to provide an exact dollar amount for OOP costs. For these respondents, the HRS poses a series of questions to determine an upper and lower limit for OOP costs, thereby bracketing the estimated spending. We used imputation methods that are standard for the HRS to impute these values (eSupplement, Supplemental Digital Content 1, http://links.lww.com/CCM/E465) (21). We adjusted costs for inflation using the Consumer Price Index for medical care and report amounts in 2014 dollars.
Based on age and reported insurance status at time of death, we categorized decedents into one of seven mutually exclusive age and insurance coverage categories: less than 65 with Medicaid or any health plan for active duty and retired members of the uniformed services, their families, and survivors (heretofore referred to as “Medicaid coverage”); less than 65 with private insurance; greater than or equal to 65 with Medicare FFS only; greater than or equal to 65 with Medicare Advantage only; greater than or equal to 65 with Medicare and Medicaid or a military health plan; greater than or equal to 65 with Medicare and private insurance; or no insurance. We pool Medicaid and military insurance due to their relatively generous coverage of long-term services. Private insurance included any coverage through an employer or business, coverage for retirees, and any Medigap or supplemental coverage. Using reported insurance status from the interview wave prior to death, we subdivided decedents in Medicaid categories into those with Medicaid coverage in both periods and those who initiated coverage between the prior interview and death (e.g., transitioned to Medicaid by spending down).
We modeled the probability of having any OOP costs in each cost category by insurance coverage using logistic regression. We modeled OOP costs using a two-part model with logistic regression and a generalized linear model with gamma family and log link. We used recycled predictions to generate predicted probabilities of having any OOP costs and predicted OOP costs combining the estimates from each part of the two-part model (eSupplement, Supplemental Digital Content 1, http://links.lww.com/CCM/E465).
Regression model covariates were obtained from the last interview wave prior to death and included age, gender, race, education, marital status, veteran status, self-rated health, number of activities of daily living limitations, number of chronic conditions, inflation-adjusted income and assets, costs incurred post-Medicare Part D implementation, and use of dialysis and mechanical ventilation during the ICU stay. To account for a recall period of varying length, the number of months in the recall period was included as a variable in the model. Predictions were then made as though all subjects were reporting 12 months of costs. We do not apply population weights since we are pooling decedents over multiple years. All analyses were conducted in Stata 14.1 (StataCorp LP, College Station, TX). The use of Medicare and HRS data were approved by the Institutional Review Board at the University of Washington, the HRS Restricted Data Applications Processing Center, and the Centers for Medicare and Medicaid Services Privacy Board.
This study included 2,909 participants; mean age was 78 (sd, 10). At death, nearly 40% had insurance coverage from both Medicare and a private plan and a quarter had Medicaid and Medicare (dual-eligible). Remaining decedents with insurance coverage had either standalone Medicare FFS (13%) or Medicare Advantage plan (11%), standalone private plan (4%), or Medicaid (7%); 2% were uninsured. (Additional characteristics in Table 1).
Total unadjusted OOP costs as a percentage of total household assets are shown in Figure 1. Greater than one-fifth of insured decedents had costs totaling more than 10% of assets; 8–22% of decedents, depending on insurance coverage, had costs totaling more than 50% of assets. More than 25% of decedents with Medicaid coverage had no household assets. Consequently, any OOP medical costs would constitute 100% of assets for these households.
Results of our multivariate analyses are provided in Tables 2 and 3. Among insured decedents greater than or equal to 65, the dual-eligible had the lowest predicted probabilities of having any OOP costs in the last year of life in every cost category. (Table 2) Similarly, among decedents less than 65, those with Medicaid coverage had the lowest predicted probabilities of having any OOP costs for all categories except nursing home care. Notably, the predicted probabilities of having any OOP costs in the last year of life for nursing home care were similar regardless of age and insurance coverage.
Among insured decedents greater than or equal to 65, those with only Medicare FFS had the highest estimated OOP costs in the last year of life in every cost category except for prescription drugs. (Table 3; pairwise comparisons in eTable 1, Supplemental Digital Content 1, http://links.lww.com/CCM/E465) Total OOP costs in the year before death for this group were estimated at $12,668 (95% CI, $9,744–15,592), with hospital (likely due to repeated admissions; eTable 2, Supplemental Digital Content 1, http://links.lww.com/CCM/E465) and nursing home care as the largest cost contributors. Estimated total OOP costs for dual-eligible decedents were $5,507 (95% CI, $3,964–7,050). Total estimated OOP costs for decedents less than 65 who were covered by a private plan were not different from those less than 65 covered by Medicaid. Uninsured decedents had the highest estimated total OOP costs at $26,993 (95% CI, $6,834–47,153).
In each cost category, OOP costs in the last year of life for decedents greater than or equal to 65 covered continuously by Medicaid were far lower than for decedents who transitioned into Medicaid coverage (eTable 3, Supplemental Digital Content 1, http://links.lww.com/CCM/E465). Total OOP costs for continuously covered decedents were estimated at $2,409 (95% CI, $1,709–3,108), whereas total costs for decedents who transitioned into Medicaid were more than 4x larger at $10,045 (95% CI, $6,534–13,556).
We repeated the analyses within the subgroup of decedents who had been mechanically ventilated or on dialysis during their ICU stay (n = 1,482). Our findings were similar (eTables 4 and 5, Supplemental Digital Content 1, http://links.lww.com/CCM/E465; and eFig. 1, Supplemental Digital Content 2, http://links.lww.com/CCM/E466; legend, Supplemental Digital Content 1, http://links.lww.com/CCM/E465).
Our main findings were: 1) although designed for hospital coverage, Medicare FFS beneficiaries have significant OOP costs in the last year of life-related to hospital admissions; 2) individuals who transitioned to Medicaid coverage have almost the same OOP spending as those on Medicare alone; 3) despite the passage of Medicare Part D, OOP spending on prescription drugs exhibits little variance across insurance types except for Medicaid; and 4) nursing home use is a large contributor to OOP costs in this population.
To our knowledge, this is the first study to focus on OOP costs across different categories of insurance coverage for individuals requiring intensive care within the last months of life. Prior studies examining end-of-life financial risk limited their sample to elderly Medicare beneficiaries in the last 5 years of life, with comparisons made by cause of death (15, 16). By contrast, we identified our sample based on the receipt of high-intensity care in the last months of life; ICU patients generally have high acuity of illness requiring complex, costly care that can potentially impose greater financial burden (22). Our sample had an average of 3.5 chronic conditions with over 50% receiving life support at some time during the last months of life. We also included all ages in HRS to compare OOP costs across different age and insurance categories. In the United States, 4–6 million patients are admitted to ICUs annually, and while more than half of these patients are greater than or equal to 65, many are less than 65 and thus not Medicare eligible (22). Additionally, use of critical care services for the Medicaid population has increased disproportionately over time (3).
We found that the burden of OOP expenditures exhibits tremendous heterogeneity, both across and within health insurance coverage categories. Under the Affordable Care Act, people under 30 or those who qualify for a “hardship exemption,” are eligible for catastrophic coverage. These plans have a deductible of $7,150, after which the insurance company pays for all covered services with no copayment or coinsurance (23). With the exception of those covered by Medicare and Medicaid (dual-eligible), patients covered under every category of insurance had total OOP costs near the end of life that exceeded this threshold.
We also found that more than 20% of decedents had costs totaling more than 10% of assets, and between eight and 22% of decedents, depending on insurance, had costs totaling more than 50% of assets in the last months of life. A prior study on healthcare and wealth depletion among older adults considered a 10% depletion in household assets to be a meaningful change (24). We chose to examine percent of household assets rather than income for two main reasons: 1) household income is much less variable in older age due to social security benefits being the principal source of income for the majority of older Americans and 2) at the end of life, a person could feasibly have all of their assets available to spend on healthcare costs.
Primarily designed for hospital coverage, we expected Medicare FFS beneficiaries to be relatively insulated from high OOP hospital costs through Part A coverage. Our results, however, suggest otherwise. Among FFS beneficiaries greater than or equal to 65, average OOP costs for hospital admissions was roughly $5,000 in the last year of life, nearly double compared with those with Medicare Advantage or other supplemental coverage. Our sample of Medicare FFS beneficiaries had an average of 3 (sd, 3) hospital stays within the last year of life (eTable 2, Supplemental Digital Content 1, http://links.lww.com/CCM/E465). FFS beneficiaries pay a $1,300 co-pay for each hospital admission, suggesting that multiple admissions were the largest contributor to the observed OOP hospital costs. With no cap on OOP costs, FFS beneficiaries with no other insurance have limited protection from truly catastrophic costs (25).
Being able to separately examine individuals who transitioned to Medicaid coverage at the end of life by spending down and those who have Medicaid for a longer duration highlights the important dual-role Medicaid plays in the social safety net (26). As a supplementary insurance program for the gaps in Medicare coverage, dual-eligible elderly have the most complete hospital coverage of all the insurance groups, as demonstrated by the lowest OOP costs. Although Medicaid and military insurance are the only public insurance programs significantly financing long-term care (26), it appears that gaps in nursing home coverage remain for individuals covered by these programs. For elderly patients who are dual-eligible, likely covered for the entire period, nursing homes comprise over half of the total OOP costs.
The monthly OOP costs for prescription drugs are remarkably stable across insurance categories, including the uninsured, except for individuals covered by Medicaid (27). The spending amounts could be masking differential access to prescription drugs, or it could be showing that insurance, other than Medicaid, does surprisingly little in insulating this cohort from the financial costs of their prescriptions. Further work is needed as we debate the closure of the “donut hole” in Part D coverage and policy responses to high prescription drug prices (28).
Last, across all categories of insurance coverage, nursing home use remains a disproportionately large contributor to OOP costs. As our population continues to age and more patients survive critical illness (29–32), financing for post-acute and long-term care in the United States will remain a growing public health issue.
This study has several limitations. First, both ICU status and OOP costs were obtained via self-report and subject to recall bias. We do not expect, however, that recall would differ according to our predictor of interest, type of insurance. Second, exact OOP amounts were missing for a significant portion of the sample, requiring imputation. This limitation is not unique to our study; we used methods designed and validated for HRS to impute the data (21). An examination of the quality of the OOP medical expenditure measures in HRS concluded that average OOP costs in the HRS, when using imputed amounts, have no systematic bias and are very similar in magnitude to those reported in the Medicare Current Beneficiary Survey and the Medical Expenditure Panel Survey, two surveys considered to provide high-quality OOP cost data (33). Third, OOP cost estimates for uninsured subjects and those less than 65 with private insurance may be imprecise because of limited sample sizes. Similarly, our estimates for nursing home and other medical costs may be imprecise due to limited sample sizes.
Across all categories of insurance coverage, OOP spending in the last 12 months of life for patients requiring intensive care is high, representing a significant portion of assets for many. With the lack of an OOP maximum and a relatively high co-payment for hospitalizations, traditional Medicare FFS alone does not insulate individuals from the financial burden of high-intensity care and repeated hospitalizations. Seriously ill patients report that having healthcare costs covered to avoid placing a financial burden on loved ones is a top concern (34, 35). Increased financial burden of care has been shown to be strongly correlated with decreased health-related quality of life, indicating that it is a clinically relevant patient-centered outcome (36). The prevalence of finance-related stress, to which OOP costs contribute, is high among patients requiring intensive care and their families and is associated with symptoms of anxiety and depression (37). Clinicians have traditionally avoided discussions of healthcare costs; however, in an era where spending on healthcare is increasingly being shifted to the patient, the interdisciplinary clinician team may need to address this aspect of care for the well-being of patients, their families, and society (38). The first step in preparing for these discussions is understanding the magnitude of OOP costs patients and families may face. Our findings provide estimates on OOP expenditures for direct medical care that decedents who used the ICU near the end of life faced in the last 12 months of life, by type of care and by insurance coverage.
1. Teno JM, Gozalo P, Trivedi AN, et al. Site of death, place of care, and health care transitions among US Medicare beneficiaries, 2000–2015. JAMA 2018; 320:264–271
2. Teno JM, Gozalo PL, Bynum JP, et al. Change in end-of-life care
for Medicare beneficiaries: Site of death, place of care, and health care transitions in 2000, 2005, and 2009. JAMA 2013; 309:470–477
3. Halpern NA, Goldman DA, Tan KS, et al. Trends in critical care beds and use among population groups and Medicare and Medicaid beneficiaries in the United States: 2000–2010. Crit Care Med 2016; 44:1490–1499
5. Finkelstein A, Taubman S, Wright B, et al.; Oregon Health Study Group: The Oregon Health Insurance Experiment: Evidence from the first year. Q J Econ 2012; 127:1057–1106
6. Baicker K, Taubman SL, Allen HL, et al.; Oregon Health Study Group: The Oregon experiment–effects of Medicaid on clinical outcomes. N Engl J Med 2013; 368:1713–1722
7. Mazumder B, Miller S. The effects of the Massachusetts health reform on household financial distress. Am Econ J Econ Policy 2016; 8:284–313
8. Hu L, Kaestner R, Mazumder B, et al. The Effect of the Patient Protection and Affordable Care Act Medicaid Expansions on Financial Well-Being. 2018Cambridge, MA, National Bureau of Economic Research.
9. Martin AB, Hartman M, Benson J, et al.; National Health Expenditure Accounts Team: National health spending in 2014: Faster growth driven by coverage expansion and prescription drug spending. Health Aff (Millwood) 2016; 35:150–160
11. Osborn R, Doty MM, Moulds D, et al. Older Americans were sicker and faced more financial barriers to health care than counterparts in other countries. Health Aff (Millwood) 2017; 36:2123–2132
12. French E, Jones JB. On the distribution and dynamics of health care costs. J Appl Economet 2004; 19:705–721
13. De Nardi M, French E, Jones JB. Why do the elderly save? The role of medical expenses. J Polit Econ 2010; 19:39–75
14. Dobkin C, Finkelstein A, Kluender R, et al. The Economic Consequences of Hospital Admissions. Am Econ Rev 2018; 108:308–352
15. Kelley AS, McGarry K, Fahle S, et al. Out-of-pocket spending in the last five years of life. J Gen Intern Med 2013; 28:304–309
16. Kelley AS, McGarry K, Gorges R, et al. The burden of health care costs for patients with dementia in the last 5 years of life. Ann Intern Med 2015; 163:729–736
17. Angus DC, Barnato AE, Linde-Zwirble WT, et al.; Robert Wood Johnson Foundation ICU End-Of-Life Peer Group: Use of intensive care
at the end of life in the United States: An epidemiologic study. Crit Care Med 2004; 32:638–643
18. Sonnega A, Faul JD, Ofstedal MB, et al. Cohort profile: The Health and Retirement Study (HRS). Int J Epidemiol 2014; 43:576–585
19. Servais A. Overview of HRS Public Data Files for Cross-Sectional and Longitudinal Analysis. 2004. Ann Arbor, MI, University of Michigan Survey Research Center, Institute for Social Research. Available at: https://hrs.isr.umich.edu/documentation/user-guides
. Accessed March 12, 2019
20. Raina P, Torrance-Rynard V, Wong M, et al. Agreement between self-reported and routinely collected health-care utilization data among seniors. Health Serv Res 2002; 37:751–774
22. Halpern NA, Pastores SM, Oropello JM, et al. Critical care medicine in the United States: Addressing the intensivist shortage and image of the specialty. Crit Care Med 2013; 41:2754–2761
23. Department of Health and Human Services: Patient Protection and Affordable Care Act; health insurance market rules. Final rule. Fed Regist 2013; 78:13405–13442
24. Lee J, Kim H. An examination of the impact of health on wealth depletion in elderly individuals. J Gerontol B Psychol Sci Soc Sci 2003; 58:S120–S126
26. Rowland D, Garfield R. Health care for the poor: Medicaid at 35. Health Care Financ Rev 2000; 22:23–34
27. Kesselheim AS, Huybrechts KF, Choudhry NK, et al. Prescription drug insurance coverage and patient health outcomes: A systematic review. Am J Public Health 2015; 105:e17–e30
29. Erickson SE, Martin GS, Davis JL, et al. Recent trends in acute lung injury mortality: 1996–2005. Crit Care Med 2009; 37:1574–1579
30. Friedman G, Silva E, Vincent JL. Has the mortality of septic shock changed with time. Crit Care Med 1998; 26:2078–2086
31. Yealy DM, Kellum JA, Huang DT, et al.; ProCESS Investigators: A randomized trial of protocol-based care for early septic shock. N Engl J Med 2014; 370:1683–1693
32. Zimmerman JE, Kramer AA, Knaus WA. Changes in hospital mortality for United States intensive care
unit admissions from 1988 to 2012. Crit Care 2013; 17:R81
33. Goldman DP, Zissimopoulos J, Lu Y. Medical expenditure measures in the health and retirement study. Forum Health Econ Policy 2011; 14:4
34. Downey L, Engelberg RA, Curtis JR, et al. Shared priorities for the end-of-life period. J Pain Symptom Manage 2009; 37:175–188
35. Steinhauser KE, Christakis NA, Clipp EC, et al. Factors considered important at the end of life by patients, family, physicians, and other care providers. JAMA 2000; 284:2476–2482
36. de Souza JA, Yap BJ, Wroblewski K, et al. Measuring financial toxicity as a clinically relevant patient-reported outcome: The validation of the COmprehensive Score for financial Toxicity (COST). Cancer 2017; 123:476–484
37. Khandelwal N, Hough CL, Downey L, et al. Prevalence, risk factors, and outcomes of financial stress in survivors of critical illness. Crit Care Med 2018; 46:e530–e539
38. Moriates C, Shah NT, Arora VM. First, do no (financial) harm. JAMA 2013; 310:577–578