SRDR+

1. Review ID

(surname of first author and year first full report of study was published e.g. Smith 2001)


	Okun 2013

2. Review title


	Volunteering by Older Adults and Risk of Mortality: A Meta-Analysis

3. Date form completed


	09/08/2022

4. Initials of person extracting


	BN

5. Review funding source


	No information

6. Possible conflicts of interest


	No information

7. Aim of review


	Extending previous reviews, we carried out a meta-analysis of the relation between organizational volunteering by late-middle-aged and older adults (minimum age ⫽ 55 years old) and risk of mortality.

8. Number of databases searched


	2

9. Names of databases searched; date ranges of databases searched


	Medline and PsycINFO

10. Date of last search


	November 3, 2011

11. Number of included studies


	13

12. Exclusion criteria for participants

(e.g age, comorbidities) 


	excluded: younger adults, only older adults included

13. Exclusion criteria for volunteering

(e.g type of volunteering, for a specific organistion/purpose)


	Organisational volunteering

14. Exclusion criteria for study type


	journal article or book chapter written in English reporting on empirical research, prospective studies

15. Exclusion criteria for outcome measures


	mortality (on an individual level)

16. Outcomes studied

(select all that apply)


	Psychological Physical Social General

17. Primary reported outcomes


	Mortality

18. Secondary reported outcomes (if applicable)


	N/A

19. Number of participants included in the review


	The total sample sizes, which do not necessarily correspond to the sample sizes associated with the effect sizes, ranged from 868 to 15,938, with a median of 4,927.50.

20. Review’s included study type (% of quant studies)


	100% quant, all prospective: The mean and standard deviation for length of the mortality surveillance period in years were 5.94 and 1.86, respectively.

21. Included studies countries of publication


	Nine of the studies used U.S. samples, and the remaining three studies employed Israeli (n ⫽ 2) and Taiwanese samples.

22. Range of included studies years of publication


	The articles were published over a span exceeding 25 years. Four articles were published prior to 2000, seven articles were published between 2000 and 2009, and the remaining three articles were published between 2010 and 2012.

23. Review’s population

(age, ethnicity, SES)


	The minimum age of the participants ranged from 55 to 75 years old, with a median of 66.50 years old.

24. Social outcomes reported

25. Social outcomes not supported

(e.g cited as non-significant)

26. Physical outcomes reported

27. Physical outcomes not supported

(e.g cited as non-significant)

28. Psychological outcomes reported

29. Psychological outcomes not supported

(e.g cited as non-significant)

30. General outcomes reported

(i.e general health and wellbeing)


	in the absence of control variables, the average effect size suggests that relative to nonvolunteers, volunteers have a 47% decrease in the risk of death, with a 95% confidence interval of 38% to 55% (25). in the presence of control variables (most often SES, health behaviours, and marital status), the average effect size indicates that, relative to nonvolunteers, volunteers have a 24% decrease in the risk of death, with a 95% confidence interval of 16% to 31%.

31. General outcomes not supported

(e.g cited as non-significant) 

32. Interactions reported

(i.e between each other or demographic variables)


	To examine the reduction in the relation between volunteering and mortality associated with the introduction of control variables, we used matched pairs of effect sizes from the nine studies that yielded both adjusted and unadjusted effect sizes. the difference between the adjusted and unadjusted effect sizes ranged from 0.07 to 0.32. the HR, on average, increases by 0.20 when the adjusted and unadjusted effect sizes are directly compared across the set of nine studies, meaning the magnitude of the relation between volunteering and mortality risk is significantly (p < .001) reduced by the inclusion of covariates. Investigated whether the relationwas linear or not. of the 10 comparison, only 2 were statistically significant, both in different directions (one supporting linear, one supporting curvilinear). The results of these analyses do not provide a warrant for drawing a firm conclusion regarding whether the volunteering–mortality risk association is linear or curvilinear. No significant influence of the study country, journal impact factor, year of publication, minimum age of the sample, % deceased, % of the sample volunteering, or whether the focus of the study was on volunteering -> mortality or not (11). the adjusted relation between volunteering and mortality risk was inverse among Taiwanese men (0.81) but positive and stronger among Taiwanese women (2.28) (1). the unadjusted relation between volunteering and mortality risk was stronger among participants who were primarily motivated to volunteer by concerns for others as opposed to concerns for self. (1). Religiosity (3) and social connection (3) and social interaction (1) were both supported as moderators to the relationship. Mixed results for leisure as a moderator- yes (1) and no (1).

To examine the reduction in the relation between volunteering
and mortality associated with the introduction of control variables, we used matched pairs of effect sizes from the nine studies that yielded both adjusted and unadjusted effect sizes. the difference between the adjusted and unadjusted effect sizes ranged from 0.07 to 0.32. the HR, on average, increases by 0.20 when the adjusted and unadjusted effect sizes are directly compared across the set of nine studies, meaning the magnitude of the relation between volunteering and mortality risk is significantly (p < .001) reduced by the inclusion of covariates.
Investigated whether the relationwas linear or not. of the 10 comparison, only 2 were statistically significant, both in different directions (one supporting linear, one supporting curvilinear). The results of these analyses do not provide a warrant for drawing a firm conclusion regarding whether the volunteering–mortality risk association is linear or curvilinear.
No significant influence of the study country, journal impact factor, year of publication, minimum age of the sample, % deceased, % of the sample volunteering, or whether the focus of the study was on volunteering -> mortality or not (11). 
the adjusted relation between volunteering and mortality risk was inverse among Taiwanese men (0.81) but positive and stronger among Taiwanese women (2.28) (1).
 the unadjusted relation between volunteering and mortality risk was stronger among participants who were primarily motivated to volunteer by concerns for others as opposed to concerns for self. (1).
Religiosity (3) and social connection (3) and social interaction (1) were both supported as moderators to the relationship.
Mixed results for leisure as a moderator- yes (1) and no (1).

33. Was a meta-analysis performed?

34. Number of included studies in the meta-analysis


	Unadjusted: 25 effect sizes Adjusted: 11

35. Heterogeneity

(e.g I squared)


	I squared Unadjusted (no control variables): 82% Adjusted: 59%

36. Pooled estimates


	Unadjusted (no control variables): .53, p < .001 Adjusted: .76, p < .001

37. Confidence intervals (95%)


	Unadjusted (no control variables): 0.45 to 0.62 Adjusted: 0.69 to 0.84

38. Key conclusions from study authors


	Across 11 studies, volunteerism appeared to reduce mortality risk by almost half in unadjusted models when variables that likely mediate the effect are not first removed from the analysis. When the more conservative test is applied, one that controls for covariates such as age, sex, ethnicity, socioeconomic status, work status, marital status, religiosity, emotional health, health behaviors, social connection, social interaction, and physical health, the adjusted effect size remains substantial, predicting a 25% reduction in the risk of death. Furthermore, we detected no evidence of publication bias for our estimates of the means of the unadjusted and adjusted effect sizes. However, researchers did exhibit a bias toward reporting interaction effect sizes only when the tests of the interaction effects reached conventional levels of statistical significance. Keeping this caveat in mind, our analyses revealed that religious involvement appears to amplify the association between volunteering and mortality risk.

Across 11 studies, volunteerism appeared to reduce mortality risk by almost half in unadjusted models when variables that likely mediate the effect are not first removed from the analysis. When the more conservative test is applied, one that controls for covariates such as age, sex, ethnicity, socioeconomic status, work status, marital status, religiosity, emotional health, health behaviors, social connection, social interaction, and physical health, the adjusted effect size remains substantial, predicting a 25% reduction in the risk of death. Furthermore, we detected no evidence of publication bias for our estimates of the means of the unadjusted and adjusted effect sizes. However, researchers did exhibit a bias toward reporting interaction effect sizes only when the tests of the interaction effects reached conventional levels of statistical significance. Keeping this caveat in mind, our analyses revealed that religious involvement appears to amplify the association between volunteering and mortality risk.

39. Review limitations


	it was limited by the small, published literature and a relatively few “file draw” studies with null findings would negate the associations that we observed between volunteering and mortality risk. Second, the studies used nonexperimental designs. Although researchers, on average, controlled for more than nine types of covariates, these efforts do not permit us to draw conclusions regarding the causal impact of volunteering on mortality. lack of standardization of volunteer predictor variables.

40. AMSTAR 2 quality appraisal rating


	-9

41. Quality appraisal tool used by review (if applicable)


	None

42. Quality of included studies (if applicable)


	N/A

43. Publication bias reported (if applicable)


	trim and fill procedure: Application of the trim and fill procedure separately for the unadjusted effect sizes and the adjusted effect sizes revealed that the means of the distributions of effect sizes did not change, indicating the absence of publication bias. Was evidence for bias in reporting of moderators: interaction effect sizes included in the current meta-analysis overestimate the magnitude of the joint effect of volunteering and moderator variables on mortality risk.

44. Was correspondence required for further study information?

45. What further correspondence was required, and from whom?

46. What further study information was requested (from whom, what and when)?

47. What correspondence was received (from whom, what and when)?

Extraction form for project: The effect of volunteering on the health and wellbeing of volunteers: an umbrella review

Design Details