Study selection
A total of 1,995 studies were retrieved from the three electronic databases. After removing duplicates, 1,233 references were screened, with 118 selected for full-text review. Eight studies were included in the systematic review (Appendix 2). The most common reason for exclusion was that the sample population did not meet the inclusion criteria. Details on the exclusion process are described in the PRISMA flowchart (Fig. 1). Two independent reviewers completed data extraction.
Figure 1: PRISMA Flowchart of Screening and Review Process.
*** Fig. 1 about here ***
Characteristics of included studies
The included studies were published between 1997 and 2022. Four studies used national samples [2, 3, 5, 6] and the remaining four used provincial samples from Alberta or Ontario [1, 4, 7, 8]. The sample size of the selected studies ranged from 32 participants [8] to 2009 participants [4]. Recruitment methods primarily involved sending letters via post or emails to randomly selected residents. Six studies stratified their samples by age and sex [2–6, 8]. One study stratified by region [2], while another included stratification by income and education levels, along with age and sex [8]. Characteristics of the included studies are presented in Table 1.
Table 1
Characteristics of Included Studies.
| Authors (Year) | Sample size | Sample population | Participant sampling and recruitment | Method of administration | Study design | Single or repeated surveys | Trade-offs investigated |
| Choudhry et al. (1997) | 80 | Ontario | 135 most senior officials in the Ontario Ministry of Health (MHO) were contacted through the post. | Postal | Cross-sectional study | Single | LE, NT, and gain in LE |
| Denburg et al. (2020) | 1556 | Pan-Canadian | Emails sent to a random sample of residents who were part of an existing online panel. Recruitment based on age-sex stratification. | Online | Experimental study, with the respondents in the experimental arm were randomly assigned to a moral reasoning exercise prior to the choice scenarios | Single | LE |
| Dragojlovic et al. (2015) | 2005 | Pan-Canadian | Canadians aged 19 years and over recruited through a research company (IPSOS Reid Canada). The mode of recruitment communication was not mentioned. | Online | Experimental study - Half of the respondents were randomly assigned to one of two versions of the study: ‘Extra Funds’, i.e. allocation of new funds received from the MHO, and ‘Existing Funds’, i.e. allocation of current dollars | Single | NT |
| Hurley et al. (2020) | 1964 | Ontario | Letters sent in the post to community-dwelling adults Recruitment based on age-sex stratification | Online | Cross-sectional study | Single | HALE and Income |
| Johri et al. (2008) | 2009 | Pan-Canadian | Emails sent to residents who were part of an existing online panel. Recruitment based on age-sex stratification. | Online | Experimental study with a baseline and follow-up survey after 7 weeks. Respondents in the experimental arm were randomly assigned to a moral reasoning exercise after the choice scenarios | Repeated after 7 weeks | LE |
| Skedgel et al. (2014) | 738 | Multiple Canadian provinces: Nova Scotia, Ontario, British Columbia | Two groups were included in this survey: (1) A sample drawn from an existing online research panel. (2) A sample of decision-making agents invited to participate by emails and flyers. Recruitment for sample (1) was age-sex stratified. | Online | Cross-sectional study | Single | LE, NT and gain LE |
| Spackman et al. (2022) | 574 | Alberta | Probability sampling; contacted between May and July 2021 | Online | Cross-sectional study | Single | LE, gain in LE, QoL improvements |
| Stafinski et al. (2017) | 32 | Alberta (Edmonton and Calgary) | Letters sent in the post to the residents of Calgary and Edmonton. Recruitment based on age-sex, region, education level, and household income stratification. | In-person discussion | Cross-sectional study | Repeated the same day | NT, age, current health, LE without treatment, dependents |
| HALE: Health-adjusted life years; LE: life-expectancy at birth; NT: number treated, MHO: Ontario Ministry of Health. |
Design of questionnaire experiments
Details of the study designs are included in Table 2. All studies focused on trade-off scenarios where participants were asked to allocate resources to different population groups. One study focused on allocating resources to different socioeconomic groups [4], four looked at age-related allocations [1, 2, 5, 8], two on rare or common diseases [3, 7], and one study on cancer [6]. Other studies addressed multiple factors including disease severity, life expectancy and quality, duration of disease, and the number of individuals treated [1, 6, 7, 8]. In terms of measurement, one study estimated inequality aversion concerning income and health distributions [4]. The others used mean preference scores [2, 5], percentage preferences [1, 3], compensating variation [6], rate of agreement between respondent groups [8], and effect of equity domains on participant utility [7].
Table 2
Design of the included studies
| Authors (date) | Focus of Relevant Questions in the Study | Range of Relevant Trade-Off | Description of Survey Scenario | Were Participants Asked to Choose an Allocation Principle? |
| Choudhry et al. (1997) | - Trade-off between allocating large health gains to a few people vs. small gains to many - Trade-off between allocating health gains to younger versus older patients | LE: 30 vs. 50 years, 5 vs. 65 years NT: 500 to 10,000 Gain in LE: 1–20 years | - Participants chose between two hypothetical programs - Scenarios varied in terms of the number of people affected, average survival benefit, side-effects/harms, and the average age of patients - Scenarios focused on the distribution of benefits and age effects | No – there was no indication that participants had to explain their reasoning or choose an allocation principle. |
| Denburg et al. (2020) | - Trade-off between allocating health gains to children versus adults for hypothetical treatment scenarios included chronic blood disease, liver transplant, cancer therapy, palliative care, and eating disorders | Treatment allocation for patients aged 10 vs 40. Preferences on a sliding scale from − 5 (full preference for children) to + 5 (full preference for adults) | - Participants were asked to decide between two treatments to fund - One drug treats the child form of the disease while the other treats the adult form of the disease - Half of the participants completed a moral reasoning exercise prior to each scenario | Yes – participants were invited to select an allocation principle: Equal treatment; Relief pain and suffering; At risk of dying; Capacity to benefit longer; Most vulnerable; Evidence that it works; Live a full life; Treat those dependent on others; Family responsibility; Productive people; Special people; Rare disease |
| Dragojlovic et al. (2015) | - Trade-off between allocating health gains to common disease versus rare disease patients | Rare disease (100 patients) vs common disease (10,000 cases) | - Participants were asked to allocate funding between (1) rare disease and common disease patients; and (2) rare disease patients and other competing healthcare/non-healthcare options | No – there was no indication that participants had to choose an allocation principle |
| Hurley et al. (2020) | - Three trade-off scenarios involving allocation of income, health and income-related health | Income distribution, varying between $12,200 and $168,500 Health-adjusted life expectancy (HALE) varying between 55 and 88 years | - Participants were asked to choose between two policy options - Policy A is more equitable, whereas policy B reflects health maximization | Yes – Most (90%) participants provided reasoning for choosing one policy over the other. The most common rationales were concerns for inequality and the worst-off. The other rationales included the opportunity for higher outcomes and higher mean. |
| Johri et al. (2008) | - Trade-off between allocating health gains to younger versus older patients through life-saving programs (i.e. liver and lung transplant and coronary bypass surgery), depression treatment and palliative care | Treatment allocation for patients aged 35 vs 65. Preferences on a sliding scale from − 5 (target younger patients) to + 5 (target older patients) | - Participants chose between hypothetical health programs - Scenarios were identical except for the average age of patients - Half of the respondents completed a moral reasoning exercise after the allocation scenario was introduced | Yes – participants were invited to select 3 of 10 possible allocation principles: Equal treatment; Patient need; Relief from suffering; Capacity to benefit/best outcomes; Maximise number helped; Family responsibilities; Guarantee chance for “full life”; Duration of benefit; Personal responsibility for health; Economic productivity |
| Skedgel et al. (2014) | - Trade-off between allocating health gains between two populations of cancer patients that varied in age, disease severity, mortality rate, and number of individuals treated | Age: 10, 40, 70 LE: 1 month, 5 years and 10 years NT: 100, 2500, 5000 Gain in LE: 1, 5, 10 years | - Participants were asked to decide which health care program to fund - The two programs varied in terms of age, initial utility, initial life expectancy, final utility, gain in life expectancy, patients treated | No – there was no indication that participants had to explain their reasoning/choose an allocation principle. |
| Spackman et al. (2022) | - Trade-offs between health and equity factors | Not reported | Discrete choice experiment with attributes like health provided, life-expectancy, quality-of-life, potential conflict with patient beliefs, population characteristics, average time with disease, societal treatment fairness, and rarity of disease | Yes – participants chose based on presented scenarios emphasizing health maximization, respect for patient beliefs, and equity for unfairly treated populations |
| Stafinski et al. (2017) | - Trade-off between allocating health gains between two population groups that varied in age, disease severity, mortality rate, and number of individuals treated | Age: 20–30 years vs. 60–70 years NT: few vs many Gain in LE: Few weeks vs. 2–5 years Current health status: severe, moderate or mild illness Health outcome: full, sufficient and insufficient functioning Dependents: yes, no | - Participants were grouped into discussion groups of 16 individuals - Groups must decide between two health technologies to fund - For a given scenario, the two populations differed in terms of age, current health state, health outcome with technology, prognosis without the technology, dependents, and number of patients who would benefit. | Yes – the participants had to reach a consensus as a group and provide reasoning for their decision. |
| LE life-expectancy at birth, NT number treated. |
Six studies used online questionnaires [2–7], one used a paper-based questionnaire sent by post [1], and one employed an in-person discussion method [8]. Six studies used single questionnaires [1–4, 6, 7], while the other two employed repeated questionnaires [5, 8]. Among the repeated questionnaires, one occurred on the same day [8], and the other after seven weeks [5]. Two experimental studies randomly distributed a moral reasoning exercise to half of the respondents [2, 5]; one of them had the respondents complete the exercise before each scenario [2], while the other had them complete it after each scenario description [5]. Another experimental study assigned participants to different questionnaire versions with varied framing effects [3].
*** Table 1 about here ***
*** Table 2 about here ***
Public values in preference studies
The results of preference studies are presented in Table 3 and discussed below.
Table 3
Results of included studies
| Authors (date) | Choice Context | Results | Interpretation of Results (as per the authors) |
| Choudhry et al. (1997) | Trade-offs in age and distribution of benefits | Beneficiaries differ in average age - Program A benefits 30-year-olds (preferred by 42.5%) and Program B benefits 50-year-olds (preferred by 1.3%). Remaining participants were undecide or indifferent. - Program A benefits 5-year-olds (preferred by 57.5%) and Program B benefits 65-year-olds (8.8%). Remaining participants were undecide or indifferent Beneficiaries differ in baseline distribution of benefits Program A: 500 people gain 20 years of life Program B: 10,000 people gain 1 year of life − 55.8% preferred Program A; 18.8% preferred Program B; remaining were decide/indifferent Program A: 1,000 people gain 20 years of life Program B: 4,000 people gain 5 year of life − 30.0% preferred Program A; 25.0% preferred Program B; remaining were decide/indifferent Program A: 500 people aged 30 gain 20 years of life Program B: 2,000 people aged 50 gain 5 year of life − 53.8% preferred Program A; 21.3% preferred Program B; remaining were decide/indifferent | Beneficiaries differ in average age: Participants showed stronger preferences for treating younger patients, and when the age difference increased, the proportion of undecided participants decreased. Beneficiaries differ in the size of benefits: Participants showed a preference larger benefits for fewer people (distributional preference). When the distributional difference decreased, more participants were undecided, but the relative proportion preferring one program over the other were similar. When the distributional difference was combined with a minor age difference, the proportion of undecided responses decreased again. Additionally, respondents preferred the program that provided more benefit to fewer patients when these patients were younger. |
| Denburg et al. (2020) | Trade-offs in age and specific health conditions | Mean preference scores as for the intervention (moral reasoning) and control groups. Scores ranged from − 5 (preference for younger patients) to + 5 (preference for older patients); 0 indicates no age preference. Chronic disease: Intervention = 0.25; control = -0.47 Liver transplant: Intervention = 0.05; control = -0.49 Cancer therapy: Intervention = -0.83; control = -1.77 Palliative care: Intervention = -0.02; control = -0.43 Eating disorder: Intervention = -1.11; control = -2.01 | Those in the control group preferred allocation of resources to children in all scenarios. The strongest preference was for cancer therapy and eating disorders with largest QALY gains. Those participating in the moral reasoning exercise showed a significant preference for allocation to children in the cancer therapy and eating disorder treatment scenarios but not in the chronic disease, liver transplant, and palliative care scenarios. |
| Dragojlovic et al. (2015) | Trade-offs between treating patients with rare vs common diseases | The two survey versions (i.e. Extra Funds and Existing Funds) included scenarios where cost was either presented as equal or unequal. Extra funds, equal cost 100 rare vs 100 common disease patients: - Preferred rare (29.3%); common (40.3%); rest indifferent Extra funds, unequal cost 100 rare vs 400 common disease patients: - Preferred rare (19.9%); common (56.4%) Existing funds, equal cost 100 rare vs 100 common disease patients: - Preferred rare (34.9%); common (34.7%) Existing funds, unequal cost 100 rare vs 400 common disease patients: - Preferred rare (23.2%); common (48.3%) | The majority of the respondents preferred to treat common-disease patients, and between 24–30% of respondents were indifferent. Respondents allocated to the ‘existing funds' survey were more likely to express indifference than respondents in the ‘extra funds' survey. |
| Hurley et al. (2020) | Income, health and socioeconomic-related health inequality | The mean inequality aversion estimates (Atkinson parameter) were calculated for the univariate distributions of income and health and the bivariate distribution of income-related health. These values are presented below: - Income = 3.27 - Health = 1.17 - Income-related health = 1.66 | Income: The mean aversion estimate of 3.27 for income implies that society would be willing to give up ∼56% of mean income if the remainder were distributed equally. Health: The mean aversion estimate for health is 1.17 and is non-statistically significant. Income-related health: For the bivariate income-health distribution, the aversion estimate of 1.66 implies that society would be willing to give up 3% of mean health-expected life expectancy if the remainder were distributed equally. |
| Johri et al. (2008) | Age and specific health conditions, i.e. life-saving treatment, depression treatment and palliative care | Mean preference scores for the intervention (moral reasoning) and control groups at baseline and follow-up at 7 weeks. Scores ranged from − 5 (preference for younger patients) to + 5 (preference for older patients); 0 indicates no age preference. Liver transplant (baseline and follow-up, respectively): - Intervention: -1.35 and − 1.19; control: -1.89 and − 1.48 Palliative care: - Intervention: 0.01 and − 0.05; control: 0.15 and 0.01 Depression treatment: - Intervention: -0.71 and − 0.67; control: -1.04 and − 0.92 Lung transplant: - Intervention: -1.24 and − 1.10; control: -1.86 and − 1.57 Coronary bypass (baseline and follow-up, respectively): - Intervention: -0.83 and − 0.85; control: -1.22 and − 1.07 | Participants in the intervention group (i.e. moral reasoning) showed weaker preferences for treating younger patients than those in the control group. In baseline survey, differences between the two groups were significant (p < 0.001) for all scenarios except palliative care. In the follow-up survey, differences between experimental groups were significant for three of five scenarios (lung transplant (p < 0.001), liver transplant, and depression treatment (p < 0.05)). For all groups, rates of "no preference" responses were higher for the intervention group (p < 0.001). |
| Skedgel et al. (2014) | Trade-offs in age, disease severity, and distribution of benefits | The trade-offs were measured in terms of compensating variation (CV). For each attribute, CV was estimated for upward and downward changes from the baseline (middle) level and can be interpreted as the willingness to sacrifice individual life-year gains in order to secure greater (lesser) priority for a more (less) desirable level. Patient age (10 years old, 40 years old, 70 years old): CV, Baseline to Low (10 years old) = − 4.36; Baseline to High (70 years old) = 2.91 Initial health utility (0.1, 0.5, 0.9): CV, Baseline to Low (0.1) = − 0.57; Baseline to High (0.9) = 1.41 Life expectancy (1 month, 5 years, 10 years): CV, Baseline to Low (1 month) = 3.57; Baseline to High (10 years) = − 0.77 Final health utility (0.1, 0.5, 0.9): CV, Baseline to Low (0.1) = 2.88; Baseline to High (0.9) = 0.71 Total patients treated (100, 2500, 5000): CV, Baseline to Low (100) = − 0.60; Baseline to High (5000) = − 4.20 | There was a significant preference for prioritizing younger and larger patient groups and patients with the greatest initial life expectancy. There were no significant welfare effects over the initial level of health state or the best final health state in the smallest patient groups relative to baseline. |
| Spackman et al. (2022) | Trade-offs between health and equity, including factors like life-expectancy, quality-of-life, potential conflicts with patients’ beliefs, population characteristics, average time with disease, historical unfair treatment, and rarity of disease. | Baseline life-expectancy or quality-of-life, time with disease and whether the disease was rare did not have a statistically significant effect on utility. Gain in life-expectancy increased utility by 0.21 per year gained, utility increased by 0.05 for each 0.01 improvement in quality-of-life, treatments for patients that had been treated unfairly by society increased utility by 0.09 and treatments that respected all patients’ beliefs increased utility by 0.17. | Willingness among respondents to prioritize treatments that respect patients' beliefs and address populations historically treated unfairly over strict health maximization. |
| Stafinski et al. (2017) | Trade-offs in age, disease severity, and distribution of benefits | The rate of agreement between the two Jury groups for their decision in funding population A or B in the hypothetical discussion scenarios is presented below. No other quantitative data was reported. Session 1: 100% (16/16 scenarios) Session 2: 100% (16/16 scenarios) Session 3: 96.9% (31/32 scenarios) | No quantitative analysis was conducted. Qualitative results suggested: - If population sizes were the same, there was a preference for funding the youth unless the older population is facing imminent death. - There was a preference for funding technologies that could return patients to at least sufficient functioning, regardless of other characteristics. - There was a preference for funding technologies that help those worse off as well as those facing imminent death. |
Preferences related to income and fairness
Only one study [4] examined preferences in relation to socioeconomic-related inequalities. Specifically, it asked participants about their preferred allocation of income and health across income, health and income-related health quintiles. The study presented two policy options that required participants to make a trade-off between equity and efficiency. The findings revealed that participants were willing to sacrifice ~ 56% of the mean societal income to achieve a more equitable income distribution. In contrast, there was a reluctance to sacrifice overall health to improve inequality in the distribution of health, implying that the society is neutral to socioeconomic health inequality. Interestingly, ~ 50% and ~ 20% of participants expressed weak inequality aversion or even ‘inequality-loving’ preferences in relation the health and income-related health distributions.
The study quantified inequality aversion using the Atkinson parameter, with the values for income, health and income-related health inequality aversion being 3.27, 1.17 (not statistically significant) and 1.66, respectively. The authors found heterogeneity in preferences, based on age, income, education and employment status. Over 90% of respondents provided reasoning in the open-ended space. Rationale for inequality averse responses included concern for the worse off and preference for fairness, while the rationale for less inequality averse responses included equal claim of individuals to good health and the principle of the greatest good for the greatest number.
Another study, Spackman et al (2022)16 included a discrete choice experiment involving 1,445 participants in Alberta to investigate the decision utility associated with patient and treatment attributes. The authors found that prioritizing patients who were unfairly treated by the society increased participant utility by 0.09 while treatments that respected patients’ beliefs increased utility by 0.17. However, the study did not provide an interpretation of the relative magnitude of preference values.
Preferences related to age
Four studies examined age-based preferences in healthcare resource allocation [1, 2, 5, 8]. In two studies [2, 5], a similar design was used in which half of the respondents were randomly selected to complete a moral reasoning exercise (intervention) prior to completing the survey – these studies calculated mean preference scores for the intervention and control groups by scenario. Preference scores ranged from − 5, representing the strongest preference for children/younger patients, to + 5, representing the strongest preference for adults/older patients; 0 indicated no age preference. Preference scores were used to determine which program the respondents would rather fund, i.e. one favouring children/younger patients and another favouring adults/older patients. The moral reasoning exercise exposed participants in the intervention group to the ethical principles relevant to resource allocation decisions that considered a range of ideas, including equal access to care and funding treatment based on effectiveness, age, capacity to benefit, suffering, risk of death, rarity, dependence on others, family responsibilities, personal responsibility for health, economic productivity and numbers treated. Denburg et al (2020)17 performed only a single survey comparing intervention and control results whereas, Johri et al (2008)18 repeated this study after seven weeks to compare the preference scores from baseline and follow-up surveys.
Johri et al (2008)18 included liver transplant, palliative care, depression treatment, lung transplant, and coronary bypass in their study as hypothetical resource allocation scenarios. This study found that, in all cases except palliative care, respondents preferred allocating resources to younger patients. When exposed to the moral reasoning exercise, participants preference for prioritizing patients with specific conditions over others, reduced significantly. Additionally, rates of no preference were also higher for the moral reasoning group. Denburg et al (2020)17 included chronic disease, liver transplant, cancer therapy, palliative care, and eating disorder treatment in their study as hypothetical funding scenarios, and found similar results as Johri et al, i.e. participants preferred resource allocation to children but those exposed to the moral reasoning exercise had weaker preference for children. The difference in preference between the reasoning and control groups was largest for cancer therapy and eating disorder treatment scenarios.
Choudhry et al (1997)19 surveyed senior health officials on funding preferences, finding a general preference for younger patients, especially when the benefits varied significantly. However, when the age gap decreased between scenarios, the proportion of undecided/indifferent individuals increased. When the distributional difference was large (e.g. 500 people gain 20 years of life expectancy vs. 10 000 people gain 1 year of life expectancy), respondents preferred the scenario with larger benefits for a smaller number of people suggesting evidence of a distributional preference. When the distributional difference scenario was combined with a minor age difference, respondents preferred the program that provided more benefits to fewer patients when these patients were younger.
Finally, Stafinski et al (2017)20 conducted Citizens’ Juries in Edmonton and Calgary involving 32 participants who participated in small group and plenary sessions. Participants were presented with trade-off scenarios involving age, current health state, health outcome with treatment, prognosis without treatment and dependents. No quantitative analysis was conducted other than assessing agreement rate between two Juries. There was a preference for funding the younger population unless the older patients faced imminent death and can receive the same health gain as the younger patients. Additional preferences were identified in relation to funding the worst-off, those facing imminent death and treating the larger number of patients unless the gains in the smaller group are large.
Preferences related to orphan drugs
Dragojlovic et al (2015)21 focused on resource allocation decisions involving treatment for common diseases versus those for rare diseases (i.e. orphan drugs). The cost of orphan drugs tends to be high and may not be cost-effective based on the conventional willingness-to-pay thresholds. As a result, policy makers struggle when it comes to resource allocation decisions involving rare diseases. This study surveyed 2,005 Canadian adults and included trade-offs involving treatments for rare versus common diseases. Overall, the study found that the majority of respondents preferred to fund patients with common diseases over those with rare diseases. However, ~ 27% of participants were indifferent between the two disease categories2. This study had a lower indifference rate than a similar study conducted in Norway which reported an indifference rate of ~ 65%.22 In Dragojlovic et al (2015),21 when respondents selected the indifference option, they were asked to distribute funds between common and rare disease patients; 32–48% of respondents allocated the funds equally between the two diseases. Interestingly, when indifference was not provided as a decision option, 60% of respondents chose to fund rare disease patients. However, when costs were presented as unequal (i.e. rare diseases costing four times as much as common diseases), only 30% of participants chose the rare disease option. Finally, the choice was not influenced by whether the allocated funds were considered additional money or existing. In addition to Dragojlovic (2015), Spackman et al (2022)16 also included disease rarity in their DCE design; however, this attribute did not appear to have an impact on the decision utility of participants.
Preference related to cancer patients
Skedgel et al (2014)23 developed a discrete choice experiment to evaluate participant preferences for cancer treatment in relation to the health maximization versus distributive justice. The attributes considered include patient age, disease severity, final health state, duration of benefit and distributional concerns. Each attribute was assigned three levels, and levels were evenly spaced across plausible values (e.g., for the attribute age, three levels were specified (ages 10, 40 and 70)). The trade-off was calculated as compensating variation (CV) between attributes. CV was calculated as changes from the baseline level either upwards to the high level or downwards to the low level. CV can be interpreted as the willingness to sacrifice life-years in order to give greater priority to a more desirable level or lesser priority to a lesser desirable level. The findings suggest that only 3% of respondents favored allocation based on health maximization across all choice scenarios. Participants were willing to prioritize treating younger and larger patient groups as well as patients with the greatest life expectancy. This indicates a preference for distributive justice, rather than a solely based on maximizing QALYs.
Preference related to life expectancy and quality-of-life and the size of health gains
The DCE study conducted by Spackman et al (2022)16 also included attributes related to health gains. The study found that baseline life expectancy and quality-of-life were not significantly associated with decision utility. On the other hand, treatment-related gain in life expectancy and improvement in quality-of-life increased decision utility by 0.21 per year gained and 0.05 per 0.01 improvement in quality-of-life.
Assessment of study quality
Overall, the identified studies were representative of the general Canadian or provincial population except for one study. This sample population included only seniors who were more educated than the general public [1]. The level of missing data varied across studies; one study did not present this value [3], three studies had a high/medium level of missing data [1, 2, 5] and the remainder had a low level of missing data, i.e. less than 15%. The survey response rate was only reported in five studies and varied from medium to low adequacy [1, 2, 4, 5]. The majority of studies pilot-tested their questionnaires prior to the study or referred to a previous systematic review to identify important concepts. The reliability of the survey instruments was considered high overall, with some concerns relating to the large amounts of missing data and the generally low response rate. Spackman et al (2022)16 did not include information on missing data. Details on the risk of bias assessment are provided in Appendix 3.
*** Table 3 about here ***