Table 3.
Methodology of Reviewed Studies
| Authors | Data source | Sample representativeness | Data type | Sample size | Study design | Matching or IV strategy | Methodological quality | Family effect |
|---|---|---|---|---|---|---|---|---|
| Brenna and Di Novi (2016) | SHARE, 2004–2007 (2 waves) |
Representative for the noninstitutionalized population aged 50 and older | Longitudinal | Matched treated/ Control 1,138/3,292 |
PSM | Matched on: demographics; family composition; socioeconomic variables; information on parents receiving care; self- reported probability of receiving an inheritance; mental health status and caregiver status at the first wave | Matching quality: matched on caregiver status and mental health in first wave |
Not specifically considered |
| Coe and Van Houtven (2009) | HRS, 1992–2004 (7 waves) |
Nationally representative for community-based population | Longitudinal | Sample continued caregiving = 2,557 Sample initial caregiving = 8,007 |
Simultaneous equation models (2SLS, Arellano-Bond) |
IV continued caregiving: death of mother IV initial caregiving: number of boys/girls in the household |
Strength of instrument: F-statistics: 16–837 (continued caregiving) 6–18 (initial caregiving) |
Not specifically considered |
| Di Novi and colleagues (2015) | SHARE, 2004 and 2006/2007 | Representative for the noninstitutionalized population aged 50 and older | Longitudinal | Matched treatment/ control 535/1,825 |
PSM | Matched on: socioeconomic variables; employment; family composition; occupation and income; previous SAH, CASP and caregiving status | Matching quality: Matched on caregiving status, SAH and CASP in first wave |
Not specifically considered |
| Do and colleagues (2015) | Korean LSA, 2006–2010 (3 waves) |
Nationally representative study of noninstitutionalized adults aged 45 years or older | Longitudinal | n = 2,528 (daughters-in-law) n = 4,108 (daughters) | Simultaneous equation models (2SLS, IV-probit) |
IV: ADL limitations of the mother(-in-law) and of the father(-in-law) | Strength of instrument: F-statistics: 86 (daughter- in-law) and 37 (daughter) | Aim to avoid family effect by focusing on physical health and care for parents-in-law |
| Fukahori and colleagues (2015) | Japanese panel survey on middle-aged persons, 1997–2005 | Randomly selected from the national population | Longitudinal | Matched treated/control 155/155 (males) 188/188 (spouses) |
PSM | Matched on: employment, SAH, retirement, age, education, and wage | Matching quality: Not matched on pretreatment status |
Not specifically considered |
| Goren and colleagues (2016) | Japan National Health and Wellness Surveys 2012–2013 |
Stratified by sex and age to ensure representativeness of adult population | Cross-sectional | Matched treatment/ control 1,297/1,297 |
PSM | Matched on: sex, age, BMI, exercise, alcohol, smoking, marital status, CCI (Charlson comorbidity index), insured status, education, employment, income, and children in household | Matching quality: not matched on pretreatment status |
Not specifically considered |
| Heger (2017) | SHARE, 2004–2013 (4 waves) | Representative for the noninstitutionalized population aged 50 and older | Longitudinal |
n = 3,669 (female) n = 2,752 (male) |
Simultaneous equation models | IV: Indicator of whether one parent is alive | Strength of instrument: F-statistics 18–47 |
Estimate family effect by adding health of parent as variable to model |
| Hernandez and Bigatti (2010) | HEPESE, 2000/2001 | Representativeness not discussed in the article | Longitudinal (one wave used) | Matched treatment/ control 57/57 | Direct matching | Matched on: age, gender, socioeconomic status, self-reported health, and level of acculturation | Matching quality: not matched on pretreatment status |
Not specifically considered |
| Hong and colleagues (2016) | Korea Community Health Survey, 2012–2013 | Representative of the entire community- dwelling adult population in South Korea | Cross-sectional | Matched treatment/ control 3,868/3,868 |
PSM | Matched on: age, sex, education, household income, insurance type, current smoker, current drinker, and stress level | Matching quality: not matched on pretreatment status |
Not specifically considered |
| Kenny and colleagues (2014) | HILDA, 2001–2008 | Representative sample of private Australian households | Longitudinal | Matched treatment/ control 424/424 |
PSM | Matched on: age, sex, marriage/partner, children, work hours, income, education, country of birth, chronic health condition limiting work, partner with a chronic health condition, another household member with a chronic health condition, having at least one living parent and baseline year | Matching quality: matched on baseline characteristics (pretreatment) |
Not specifically considered |
| Rosso and colleagues (2015) | Woman’s Health Initiative Clinical Trial, 1993–1998 |
Representativeness of sample not mentioned. Participants were recruited at clinical centers across the United States from 1993 to 1998 to participate in clinical trials | Longitudinal | Matched treatment/ control 2,138/3,511 |
PSM | Matched on: sociodemographic variables and health (smoking, chronic illnesses, obesity status) | Matching quality: matching on baseline characteristics (not pretreatment) |
Not specifically considered |
| Schmitz and Westphal (2015) | GSOEP, 2002–2010 |
Representative longitudinal survey of households and persons living in Germany | Longitudinal | Matched treatment/ control 1,235/29,942 |
PSM | Matched on: age of mother/father; mother/ father alive; (age) partner; number of sisters; personality traits; socioeconomic variables; health status | Matching quality: Matching on health before treatment Sample stratified by care provision at t = −1 |
Not specifically considered |
| Stroka (2014) | Techniker Krankenkasser, 2007–2009 |
Administrative data from largest statutory sickness fund in Germany | Longitudinal | Matched treatment/ control 5,696/3,125,140 (males) 7,495/2,085,946 (females) |
PSM + D-in-D | Matched on: socioeconomic variables; employment; education; work position; health status | Matching quality: matched pretreatment, at baseline only noncarers |
Not specifically considered |
| Trivedi and colleagues (2014) | BRFSS, 2009/2010 |
Nationally representative survey in the United States | Cross-sectional | Matched treatment/ control 110,514/110,514 |
PSM | Matched on: socioeconomic variables; household situation; employment, income, veteran status, immunizations within the previous year, exercise, tobacco use, self-identified physical disability, obesity status; health care access; and survey characteristics | Matching quality: not matched on pretreatment status |
Not specifically considered |
| de Zwart and colleagues (2017) | SHARE, 2004, 2006, 2010, 2013 |
Representative for the noninstitutionalized population aged 50 and older | Longitudinal | Matched treatment/ control 404/10,293 |
PSM | Matched on: socioeconomic variables; household situation; wealth; health status; health and age of spouse | Matching quality: matched on pretreatment covariates + sample stratified by care provision at t = −1 |
Not specifically considered |
Note: SHARE = Survey of Health, Ageing and Retirement Europe; HRS = Health & Retirement Study; HEPESE = Hispanic Established Populations for the Epidemiologic Study of the Elderly; HILDA = Household, Income & Labour Dynamics in Australia Survey; GSOEP = German Socio-Economic Panel; BRFSS = Behavioral Risk Factor Surveillance System; PSM = propensity score matching; 2SLS = two-stage least square; D-in-D = difference-in-difference; IV = instrumental variable.