Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Med Decis Making. Author manuscript; available in PMC 2014 February 1.
Published in final edited form as:
PMCID: PMC3606811

Evidence of spillover effects of illness among household members: EQ-5D scores from a US population sample


Health affects more than just the solitary individual who participates in health-promoting activities or suffers from an acute or chronic condition. Social network research has shown that individuals share their obesity, smoking and happiness with their family members, friends and neighbors.1 More directly, taking care of an ill or disabled individual imposes a burden on the care giver that has been well-documented, both in health effects and quality of life.26 It has also been suggested that an ill individual exerts a psychic or emotional toll on family members who care about the patient, distinct from the toll of caregiving.7,8 Less often noted is the opposite, positive benefit of care taking.9 In sum, research has clearly established that our conceptualization of health effects extends beyond one individual to include those surrounding him or her, including those physically present and those emotionally connected. The implications of this “spillover” of health to surrounding individuals include the consideration of others’ health and well-being in clinical decision making, and the inclusion of the costs and benefits of multiple individuals in analyses of health interventions. The landscape of health decision making is altered by the conceptualization of health as a family affair and important research priorities come to the fore from this perspective. This paper reports on an analysis of US population data that includes a utility-based measure of the effect of chronic illness on household members. We provide EQ-5D utility estimates of this “spillover” effect of illness on surrounding individuals.

Documented spillover effects are varied and often condition-specific, ranging from specific symptomology (e.g., anxiety10, sleep disturbance11), to psychological well-being12 and physical health.5,13 The effects have been measured in the context of a variety of specific health conditions and family or caregiving relationships, such as family caregivers of stroke patients6, spouses14, parents15, children16 and siblings17 of cancer patients18, and family members and caregivers of individuals with mental illness (e.g., bipolar disorder19, schizophrenia20, and dementia10). A separate literature has attempted to quantify the spillover effect of illness for inclusion in economic evaluation.21 Health utility is a quantitative measure of quality of life that can be used to assess patient reported outcomes and to calculate the outcome measure used in cost utility analysis, quality adjusted life years (QALYs). Health utility is a subjective measure of the value assigned to living in a particular health state, measured on a scale of 0 to 1 where 0 is the value of being dead and 1 is the value of being in full or optimal health.22 Utility measures are “preference based,” meaning they reflect individuals’ subjective preferences, or opinions, about living in a particular state of health, compared with objective measures of quality of life that report symptoms or functioning. The uniform scale on which utilities are measured allows comparison of outcomes across conditions and populations. Limitations of utility measurement methods have been noted.23

Utility measures of spillover effects offer the opportunity for integration of this component of outcomes into decision analyses and cost-effectiveness analyses. Specific situations in which spillover effects are critical inputs for decision analysis have been identified, including for measurement reasons, such as utilities for health states of very young children in which the parent proxy cannot reasonably separate their own utility from that of the child24, or for face validity reasons, such as living organ transplant in which both donor and recipient are integral to the decision.25 The study reported here estimated the spillover disutility of all chronic conditions experienced by a sample of the US population, measured using one utility instrument(the EuroQol-5D (EQ-5D)). These values will inform economic evaluations based on QALYs.



We analyzed four years of Medical Expenditure Panel Survey (MEPS) data in which EQ-5D scores were collected. MEPS is a set of household-based, in-person interview surveys of the non-institutionalized US population conducted by the Agency for Healthcare Research and Quality, collecting data at five time points over a two-year span.26,27 This study used household component data from panels beginning in 2000, 2001 and 2002, plus EQ-5D responses collected in a supplemental paper questionnaire administered once annually in 2000–2003 to all MEPS panelists active at that time. The resulting data included all family medical condition data over five reporting periods (“rounds” 1–5) and two EQ-5D administrations in each of these three panels (at approximately rounds 2 and 4; see Technical Appendix for study design diagram). The EQ-5D score represents community-perspective utility for a health state described by the respondent completing the EQ-5D instrument, meaning the values that a representative sample of the general population would assign for that described health state.22 These values are recommended for use in societal perspective cost-effectiveness analysis.22,28

Our sample consisted of 24,188 adults, representing 14,043 households and including 11,867 children (EQ-5D data were collected from adults only; children supply household medical condition data for the analysis but are not included in the sample). Individuals with missing EQ-5D data were excluded, as were individuals in households where an ill person left the household before the second administration of the EQ-5D (643 adults in 447 households). MEPS data are publicly available and unidentified and approval from applicable Institutional Review Boards was granted.


We used multivariable regression analysis to estimate the independent spillover effect of household members’ chronic conditions on an index adult’s EQ-5D score. We modeled the spillover effect in two ways for purposes of confirmation: (a) a “static” model predicting an adult’s EQ-5D score at the second administration as a function of household members’ pre-existing and newly-occurring chronic conditions; and (b) a “change-score” model predicting the change in an adult’s EQ-5D score between the first and second administration as a function of household members’ chronic conditions occurring in the intervening period. Both models controlled for other known predictors of individual utility that were available in the MEPS data (the index adult’s own chronic conditions, age (continuous), race (White), gender (female), marital status (married=1), and household income (“low household income” = annual income <US$20,000=1)), and household composition. Both models utilized the longitudinal nature of the MEPS data to predict the association between household chronic illness and individual utility.

For the “static” model, the dependent variable was the index adult’s EQ-5D score at the second administration (“EQ-5D time 2”). This variable was modeled in continuous form using Duan’s two-stage approach to account for the high frequency of 1.0 scores in the data.29 The main independent variables of interest were indicators for categories of chronic conditions among the index adult’s household members, representing conditions that could impose spillover effects on the index adult’s utility. Reported chronic conditions are coded by MEPS using three-digit International Statistical Classification of Diseases and Related Health Problems, ninth edition (ICD-9) coding (see Technical Appendix for ICD-9 code categorization). We defined chronic conditions as those with duration of 12 months or expected lasting physical impairment, with distinctions for pediatric and adult conditions, per Hwang.30 Disease categories without any reported chronic conditions or with a prevalence of less than 1% for children or adults in the sample were excluded from the regression analysis in order to maximize statistical power to detect effects. Chronic conditions were categorized as “existing” or “new” based on when the condition was first reported in MEPS: existing chronic conditions were those first reported in rounds 1 or 2 of MEPS, or within approximately the first 12 months of the panel; new chronic conditions were those first reported in round 3 of the panel, or just prior to the administration of the second EQ-5D, our dependent variable. Household members were defined as children (< 18 years of age) or adults (18+ years of age). Thus, each index adult in our analytic sample had indicator variables for existing and newly-occurring chronic conditions by ICD-9 category among child and adult household members.

For the confirmatory, “change-score” model, an EQ-5D difference score was created for each index adult (“EQ-5D difference”), defined as the EQ-5D score at the second administration minus the EQ-5D score at the first administration. This variable was truncated to 0.05 increments to create a discrete outcome with a symmetric and approximately normal distribution (albeit with a high peak at 0, indicating no change between administrations) for linear modeling. Since EQ-5D scores tend to cluster at distinct values the data can be considered approaching discrete, and our transformation makes the gaps between these distinct values more uniform. We conducted a sensitivity test by modeling the outcome with a negative binomial distribution with similar results (not presented here). This model included the same set of new and existing condition categories and control variables.

Regression analyses used general estimating equations (GEE) to control for clustering at the household level. Additional analyses tested for possible confounding of the spillover variables due to the occurrence of new chronic conditions in the index adult (as a result of the household members’ conditions) by comparing models including and excluding the index adult’s own newly-occurring conditions. These analyses yielded similar results for the spillover variables and therefore only the analyses including these variables are presented here. The primary analysis did not distinguish between one or more children or adults in a household, nor between one or more conditions in any ICD-9 category; stratified analyses investigated the spillover effect of multiple chronic conditions occurring simultaneously in the household, within adults, children, and both types of household members. Model fit was assessed for the logistic model using the Likelihood ratio test, and for the linear models with an R-squared statistic. All analyses were conducted using SAS version 9.2 (SAS Institute, Cary, North Carolina).


The MEPS sample was designed to be representative of the US adult population, with a mean age of 45 years, 54% female, 64% non-Hispanic white, and 60% married (Table 1). Half of the individuals (50%) lived in a household with at least one child under 18 years, and 89% lived with at least one other adult. Chronic conditions were present in almost one-third of the adults in the sample (31%), with 37% of adults living in a household with another adult with a chronic condition. Most commonly reported adult chronic conditions were hypertension, diabetes mellitus, joint diseases and mental disorders (primarily depression). Among children, respiratory conditions (e.g., asthma and allergies) were most common. The mean EQ-5D preference weight score among our sample was 0.86. Approximately half (46.9%) of the sample reported an EQ-5D preference weight of 1.0, and the remainder reported scores from −0.11 to 0.86.

Table 1
Sample characteristics, adults in 2000–2003 MEPS Panels with complete EQ-5D data (n=24,188)

In our “static” multivariable analysis, controlling for the respondent’s own conditions and sociodemographic and household characteristics, we found an association between lower EQ-5D scores and the presence of certain chronic conditions among household members. For the first part of the two-part model, which predicts the odds ratio of the index adult having an EQ-5D score of 1.0 versus less than 1.0, existing mental and respiratory conditions among adults and children in the household, and musculoskeletal system diseases among adults in the household, were associated with a statistically significant odds ratios (ORs): for existing mental disorders among other adults in household the OR multiplier was 0.71 (95% CI: 0.64,0.79), for existing respiratory diseases among other adults 0.85 (0.75,0.97), for existing mental disorders among children in the household 0.72 (0.62 0.82), for existing respiratory diseases among children 0.88 (0.81,0.96), and for existing musculoskeletal diseases among other adults 0.83 (0.75,0.93) (Table 2a). Some newly-occurring chronic conditions in the household predicted a further reduction in the odds ratio for the index adult’s EQ-5D score: for new mental disorders among household adults the OR multiplier was 0.79 (0.65,0.97), and among children the OR multiplier was 0.64 (0.48,0.86), while other adults’ nervous/sensory system diseases had an OR of 0.76 (0.61,0.96), and other adults’ musculoskeletal system diseases an OR of 0.78 (0.65,0.95). We found the largest effect from the index adult’s own existing and new chronic conditions, such as an OR of 0.27 for existing mental disorders in one’s self (0.24,0.30) (Table 2a). A comparative summary of the associations due to household members’ chronic conditions is presented in Figure 1. Being married and having an increase in household annual income were associated with statistically significant higher odds ratio of reporting an EQ-5D score of 1.0, while increased age, female gender, and low household annual income were significantly associated with lower odds ratio.

Figure 1
Static regression model results: stage 1 (logistic) and stage 2 (linear) models predicting association between respondent’s EQ-5D score and existing and new chronic conditions in other household members, by ICD9 category. Models are adjusted for ...
Table 2
a&b. Static, 2-stage GEE model predicting index adult’s EQ-5D score at second administration time point:

In the second part of the two-stage model, which included only adults reporting EQ-5D scores less than 1.0 (46.9% of sample; n=12,852), chronic conditions among household members were less consistently associated with a lower EQ-5D score for the index adult, with only existing mental disorders among adults or children associated with a reduction in the index adult’s EQ-5D score (adults −0.013 (95% CI: −0.0.023, −0.003); children −0.019 (−0.033, −0.005)) and musculoskeletal diseases among adults −0.012 (−0.021, −0.002). Existing neoplasms (tumors) in other adults in the household were significantly associated with an increase in EQ-5D score (0.015, (0.0004, 0.029)). Newly-occurring mental disorders and musculoskeletal diseases among adults were significantly associated with larger decrements in index adults’ EQ-5D scores (−0.021 (−0.042, −.0003) and −0.029 (−0.049, −0.009), respectively) (Table 2b). The index adult’s own existing and newly occurring chronic conditions were consistently associated with a statistically significant lower EQ-5D score, ranging from a decrement of 0.013 for existing diseases of the circulatory system (−0.02, −0.006) to 0.072 for existing mental disorders (−0.082, −0.062; Table 2b).

Our change-score model showed a significant association between an occurrence of one chronic condition category among household members and an index adult’s change in EQ-5D score between measurement periods. While controlling for changes in individuals’ own conditions, sociodemographic characteristics, household characteristics, and all other chronic conditions in the household, the occurrence of new diseases of the musculoskeletal system among other adults in the household was significantly associated with a reduction in an index adult’s EQ-5D score by 0.007 (−0.012,−0.002) over time (Table 3). Some but not all categories of new conditions occurring for the index adult himself were significantly associated with a negative change in EQ-5D score over time.

Table 3
Change-score, GEE model predicting difference in index adult’s EQ-5D scores between two administrations: EQ-5D score at second minus first administration (n=24,188)

Households with multiple other adults or multiple children with chronic conditions totaled less than 1% of the analytic sample, hence the effect of multiple chronic conditions was omitted from our analysis.


Our findings provide empiric, population-level evidence of spillover effects of certain chronic conditions onto other household members. In a representative sample of the US population, and controlling for factors known to affect health utility, adults living in households with other family members chronically ill with mental health, respiratory and musculoskeletal conditions were more likely to report deficiencies in their own health utility, and to experience lower utility scores at the occurrence of chronic illness in the household. Though these effect sizes were small, their existence in a US population sample of adults provides compelling evidence that utility losses can be attributed to spillover effects from household members’ illnesses. Our findings confirmed that while one’s own health and income are the most important contributors to EQ-5D utility scores, household members’ health can exert a small effect on one’s utility above and beyond these individual-level characteristics.

The importance of spillover effects in cost effectiveness analysis and medical decision making has been noted.3,21,3134 Evaluation of the burden of disease or the effectiveness of interventions would be incomplete without consideration of the full spectrum of effects of disease on society, including both the affected individual (i.e., patient) and others.22 The questions raised by this acknowledgement are two-fold: how to measure spillover effects and how to incorporate them into analyses.35 How to measure effects has been addressed empirically and conceptually, with somewhat less attention devoted to how to incorporate effects. Empiric estimates of the spillover utility “toll” range from a high of 0.5 for family caregivers of cognitively impaired elderly persons36, to 0.07 for parents of children with activity limitations37 and 0.06 for family caregivers of elderly persons with dementia, stroke and other diseases.38 Some studies have failed to detect any effect, including samples of caregivers of individuals with depression39, rheumatoid arthritis40, and Alzheimer’s disease.41 Our results are near the lower end of this spectrum, possibly due to our population sample. Methods of measuring spillover effects include existing utility measures as we used with the EQ-5D, and instruments designed specifically to measure spillover, such as the Carer-Qol.42 Consensus largely holds at this point that spillover effects exist and should be included in evaluations.

Methods of inclusion of spillover effects into evaluations have yet to be formally compared. The calculation of QALYs at the family level may involve the combination of individual utilities, or an entirely separate family-based quality of life weighting measure. Evidence from the literature on combining utilities may be informative in this area, suggesting the inadequacy of simple summation across or within individuals.43 Consideration of the possibility of utility values outside of the 0–1 scale, both below zero and in excess of 1.0, may suggest ways in which an individual’s utility could be a combination of his/her own value and surrounding individuals’.44 The question of integration of spillover effects into cost-utility analysis remains important to the field, particularly as the evidence for the existence of these effects and their magnitude mounts.

Our results add evidence to the empiric literature demonstrating the existence of spillover. The caveats accompanying our findings show a general bias toward the null hypothesis of no effect, so may in fact add credibility to our conclusions. While most previous spillover literature has focused on specific diseases and specific caregiver populations, our study looked at spillover effects across conditions at the population level. Thus our sample included anyone living in a household with a chronically ill individual, regardless of caretaking responsibilities or familial ties. Our data also prevented consideration of ill family members who were not physically co-located with our sample adults. Our sample was thus broad in capturing spillover effects beyond caretaking, and exclusive of potentially important caregiver relationships, both of which would serve to bias our results toward underestimation of a spillover toll.

Our measure, the EQ-5D, is widely used and provides community-perspective utility values, useful for societal-perspective cost-utility analysis. The EQ-5D has acknowledged limitations in scope and depth, however, that may result in an underestimation of the effect of one person’s illness on another.45 Spillover effects on caregivers may evade the domains included in health related quality of life, occupying an entirely different “evaluative space” beyond that inhabited by evaluations focused on patients.46 It has also been suggested that family members and caretakers may find satisfaction from caretaking or experience emotional/spiritual benefits from handling adversity.7,9 Both of these suggestions imply that the EQ-5D may not fully capture the spillover effect, regardless of direction. Our findings persisted, however, despite these limitations in measurement inherent to our data, suggesting tenacity in our results.

Our analysis used complementary and confirmatory modeling approaches to test the strength and persistence of our findings. Our static model showed a clear though small spillover association between the existence of certain chronic conditions in a household and adults’ health utility. Our change-score analysis was designed to test for the association between an incident chronic condition among household members and individuals’ utility over a short time span, and found spillover for only one category of chronic conditions. We explored adaptation to spillover through the inclusion of both newly-occurring and pre-existing conditions in our analysis, and found some suggestive evidence that new conditions had a larger spillover effect than older ones, as is seen in individual-level adaptation to disease.47 (One spurious but possibly adaptive result was seen in our static model for adult household members’ existing neoplasms for individuals with EQ-5D scores below 1.0: these cancers had a significant positive spillover effect, suggesting that this condition increases the HRQOL of surrounding individuals. This result was not seen in newly occurring cancers among household members, nor in the index adult’s own existing cancer, though could possibly be evidence of adaptation to cancer.) Model caveats include design and data issues. We controlled for individual’s own conditions to exclude the possibility that that people with chronic conditions gravitate together in households, though our results are correlations not causal effects. The lack of substantial results in our change-score analysis compared to the static model reveals possible limitations in our study design, including the short time-frame between EQ-5D measures in the MEPS survey. This interim may have been insufficient to capture changes in utility in response to the incidence of a new condition in the household. Moreover, there is some temporal uncertainty in MEPS data collection, in that each interview is conducted within a 4–5 month time span while the EQ-5D instrument is included in a supplemental questionnaire mailed during a specific span. We categorized conditions conservatively to protect against errors in data temporality. In addition, MEPS does not contain the end date for chronic conditions, and we assumed that each reported conditions lasted until the end of our analytic time frame; this assumption may over-estimate the presence of chronic conditions in each household biasing our analysis against detection of an effect. And finally, most of our results fall below the level of minimally important differences for utilities48, though our goal was not to demonstrate clinical significance but rather analytic significance. Small utility effects may be relevant for policy decisions if accumulated across a population, or compared with similarly small differences. In sum, the combination of analyses that we conducted sheds additional light on the complex question of the existence of spillover effects for chronic conditions, and the presence of an effect in different modeling approaches and under different assumptions implies a strong case for our conclusions about the importance of this factor in health utility.

In conclusion, cost-effectiveness and comparative effectiveness research are informed by these results, suggesting that benefits from interventions should be measured across the family unit of the ill individual in order to fully capture the effects of a health intervention. Costs attributable to other family members are routinely included in analyses, and recommendations suggest the inclusion of all benefits, regardless of to whom they accrue.22 For conditions where there is substantial spillover, such as children’s illnesses, consideration should be given to evaluating HRQOL data from at minimum the primary family caretaker in addition to the affected “patient.” Nearly all the spillover estimates we observed inferred a negative effect from having an ill individual in the household, though positive effects have been observed in a few instances. A remaining question is how to incorporate our spillover effects into evaluations of interventions31, which would allow for a more complete picture of benefits accrued relative to resources expended.

Supplementary Material

Appendix Fig A

Appendix Fig B

Technical Appendix


The authors are grateful for research assistance provided by Gail Strickler, and programming provided by Galina Zolotusky and Acham Gebremariam.

This work was supported by awards numbers 5R01NR011880 from the National Institute of Nursing Research and 7 K02 HS014010 from the Agency for Health Care Research and Quality, both to EW. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Health Care Research and Quality, the National Institute of Nursing Research, or the National Institutes of Health.


Preliminary results from this work were presented at the 33rd Annual Meeting of the Society for Medical Decision Making, Chicago, IL, October, 2011.


1. Smith K, Christakis DA. Social networks and health. Annu Rev Sociol. 2008;34:405–429.
2. Awad AG, Voruganti LN. The burden of schizophrenia on caregivers: a review. Pharmacoeconomics. 2008;26(2):149–162. [PubMed]
3. Brouwer WB. Too important to ignore: informal caregivers and other significant others. Pharmacoeconomics. 2006;24(1):39–41. [PubMed]
4. Etters L, Goodall D, Harrison BE. Caregiver burden among dementia patient caregivers: a review of the literature. J Am Acad Nurse Pract. 2008 Aug;20(8):423–428. [PubMed]
5. Pinquart M, Sorensen S. Correlates of physical health of informal caregivers: a meta-analysis. J Gerontol B Psychol Sci Soc Sci. 2007 Mar;62(2):P126–P137. [PubMed]
6. Rigby H, Gubitz G, Phillips S. A systematic review of caregiver burden following stroke. Int J Stroke. 2009 Aug;4(4):285–292. [PubMed]
7. Al-Janabi H, Coast J, Flynn TN. What do people value when they provide unpaid care for an older person? A meta-ethnography with interview follow-up. Soc Sci Med. 2008 Jul;67(1):111–121. [PubMed]
8. Bobinac A, van Exel NJ, Rutten FF, Brouwer WB. Health effects in significant others: separating family and care-giving effects. Med Decis Making. 2011 Mar-Apr;31(2):292–298. [PubMed]
9. Brouwer WB, van Exel NJ, van den Berg B, van den Bos GA, Koopmanschap MA. Process utility from providing informal care: the benefit of caring. Health Policy. 2005 Sep 28;74(1):85–99. [PubMed]
10. Cooper C, Balamurali TB, Livingston G. A systematic review of the prevalence and covariates of anxiety in caregivers of people with dementia. Int Psychogeriatr. 2007 Apr;19(2):175–195. [PubMed]
11. Hearson B, McClement S. Sleep disturbance in family caregivers of patients with advanced cancer. Int J Palliat Nurs. 2007 Oct;13(10):495–501. [PubMed]
12. Paul F, Rattray J. Short- and long-term impact of critical illness on relatives: literature review. J Adv Nurs. 2008 May;62(3):276–292. [PubMed]
13. Schulz R, Beach SR. Caregiving as a risk factor for mortality: the Caregiver Health Effects Study. JAMA. 1999 Dec 15;282(23):2215–2219. [PubMed]
14. Couper J, Bloch S, Love A, Macvean M, Duchesne GM, Kissane D. Psychosocial adjustment of female partners of men with prostate cancer: a review of the literature. Psychooncology. 2006 Nov;15(11):937–953. [PubMed]
15. Klassen A, Raina P, Reineking S, Dix D, Pritchard S, O'Donnell M. Developing a literature base to understand the caregiving experience of parents of children with cancer: a systematic review of factors related to parental health and well-being. Support Care Cancer. 2007 Jul;15(7):807–818. [PubMed]
16. Osborn T. The psychosocial impact of parental cancer on children and adolescents: a systematic review. Psychooncology. 2007 Feb;16(2):101–126. [PubMed]
17. O' Brien I, Duffy A, Nicholl H. Impact of childhood chronic illnesses on siblings: a literature review. Br J Nurs. 2009 Jan 13;18(22):1358, 1360–1355. Dec 10-2010. [PubMed]
18. Kitrungroter L, Cohen MZ. Quality of life of family caregivers of patients with cancer: a literature review. Oncol Nurs Forum. 2006 May;33(3):625–632. [PubMed]
19. Steele A, Maruyama N, Galynker I. Psychiatric symptoms in caregivers of patients with bipolar disorder: a review. J Affect Disord. 2010 Feb;121(1–2):10–21. [PubMed]
20. Caqueo-Urizar A, Gutierrez-Maldonado J, Miranda-Castillo C. Quality of life in caregivers of patients with schizophrenia: a literature review. Health Qual Life Outcomes. 2009;7:84. [PMC free article] [PubMed]
21. Basu A, Meltzer D. Implications of spillover effects within the family for medical cost-effectiveness analysis. J Health Econ. 2005 Jul;24(4):751–773. [PubMed]
22. Gold MR. Cost-effectiveness in Health and Medicine. New York: Oxford University Press; 1996.
23. Whitehurst DG, Bryan S. Another study showing that two preference-based measures of health-related quality of life (EQ-5D and SF-6D) are not interchangeable. But why should we expect them to be? Value Health. 2011 Jun;14(4):531–538. [PubMed]
24. Prosser LA, Bridges CB, Uyeki TM, et al. Values for preventing influenza-related morbidity and vaccine adverse events in children. Health Qual Life Outcomes. 2005;3:18. [PMC free article] [PubMed]
25. Liem YS, Kock MC, Ijzermans JN, Weimar W, Visser K, Hunink MG. Living renal donors: optimizing the imaging strategy--decision- and cost-effectiveness analysis. Radiology. 2003 Jan;226(1):53–62. [PubMed]
26. Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey. [Accessed December 12, 2011]; http://wwwmepsahrqgov/mepsweb/.
27. Ezzati-Rice T, Rohde F, Greenblatt J. Methodology report no. 22. Rockville, MD: Agency for Healthcare Research and Quality; 2008. Sample design of the Medical Expenditure Panel Survey household component, 1998–2007.
28. Miller W, Robinson LA, Lawrence RS, editors. Valuing Health for Regulatory Cost-effectiveness Analysis. Washington, D.C.: National Academies Press; 2006.
29. Duan N. A Comparison of alternative models for the demand for medical care. Santa Monica, Calif.: Rand Corporation; 1982.
30. Hwang W, Weller W, Ireys H, Anderson G. Out-of-pocket medical spending for care of chronic conditions. Health Aff (Millwood) 2001 Nov-Dec;20(6):267–278. [PubMed]
31. Davidson T, Levin LA. Is the societal approach wide enough to include relatives? Incorporating relatives' costs and effects in a cost-effectiveness analysis. Appl Health Econ Health Policy. 2010;8(1):25–35. [PubMed]
32. Bonomi AE, Boudreau DM, Fishman PA, Meenan RT, Revicki DA. Is a family equal to the sum of its parts? Estimating family-level well-being for cost-effectiveness analysis. Qual Life Res. 2005 May;14(4):1127–1133. [PubMed]
33. Basu A, Dale W, Elstein A, Meltzer D. A time tradeoff method for eliciting partner's quality of life due to patient's health states in prostate cancer. Med Decis Making. 2010 May-Jun;30(3):355–365. [PubMed]
34. van Exel J, Bobinac A, Koopmanschap M, Brouwer W. The invisible hands made visible: recognizing the value of informal care in healthcare decision-making. Expert Rev Pharmacoecon Outcomes Res. 2008 Dec;8(6):557–561. [PubMed]
35. Koopmanschap MA, van Exel JN, van den Berg B, Brouwer WB. An overview of methods and applications to value informal care in economic evaluations of healthcare. Pharmacoeconomics. 2008;26(4):269–280. [PubMed]
36. Mohide EA, Torrance GW, Streiner DL, Pringle DM, Gilbert R. Measuring the wellbeing of family caregivers using the time trade-off technique. J Clin Epidemiol. 1988;41(5):475–482. [PubMed]
37. Kuhlthau K, Kahn R, Hill KS, Gnanasekaran S, Ettner SL. The well-being of parental caregivers of children with activity limitations. Matern Child Health J. 2010 Mar;14(2):155–163. [PubMed]
38. Davidson T, Krevers B, Levin LA. In pursuit of QALY weights for relatives: empirical estimates in relatives caring for older people. Eur J Health Econ. 2008 Aug;9(3):285–292. [PubMed]
39. Sewitch MJ, McCusker J, Dendukuri N, Yaffe MJ. Depression in frail elders: impact on family caregivers. Int J Geriatr Psychiatry. 2004 Jul;19(7):655–665. [PubMed]
40. Brouwer WB, van Exel NJ, van de Berg B, Dinant HJ, Koopmanschap MA, van den Bos GA. Burden of caregiving: evidence of objective burden, subjective burden, and quality of life impacts on informal caregivers of patients with rheumatoid arthritis. Arthritis Rheum. 2004 Aug 15;51(4):570–577. [PubMed]
41. Bell CM, Araki SS, Neumann PJ. The association between caregiver burden and caregiver health-related quality of life in Alzheimer disease. Alzheimer Dis Assoc Disord. 2001 Jul-Sep;15(3):129–136. [PubMed]
42. Dixon S, Walker M, Salek S. Incorporating carer effects into economic evaluation. Pharmacoeconomics. 2006;24(1):43–53. [PubMed]
43. Fu AZ, Kattan MW. Utilities should not be multiplied: evidence from the preference-based scores in the United States. Med Care. 2008 Sep;46(9):984–990. [PubMed]
44. Sullivan PW. Are utilities bounded at 1.0? Implications for statistical analysis and scale development. Med Decis Making. 2011 Nov-Dec;31(6):787–789. [PubMed]
45. Prosser L, Grosse SD, Wittenberg E. Health utility elicitation: Is there still a role for direct methods? Pharmacoeconomics. 2012;30(2):83–86. [PubMed]
46. Al-Janabi H, Flynn TN, Coast J. QALYs and Carers. Pharmacoeconomics. 2011 Oct 19;29(12):1015–1023. [PubMed]
47. Ubel PA, Loewenstein G, Jepson C. Whose quality of life? A commentary exploring discrepancies between health state evaluations of patients and the general public. Qual Life Res. 2003 Sep;12(6):599–607. [PubMed]
48. Luo N, Johnson J, Coons SJ. Using instrument-defined health state transitions to estimate minimally important differences for four preference-based health-related quality of life instruments. Med Care. 2010 Apr;48(4):365–371. [PubMed]