Health outcomes related to the provision of free, tangible goods: A systematic review

Background Free provision of tangible goods that may improve health is one approach to addressing discrepancies in health outcomes related to income, yet it is unclear whether providing goods for free improves health. We systematically reviewed the literature that reported the association between the free provision of tangible goods and health outcomes. Methods A search was performed for relevant literature in all languages from 1995-May 2017. Eligible studies were observational and experimental which had at least one tangible item provided for free and had at least one quantitative measure of health. Studies were excluded if the intervention was primarily a service and the free good was relatively unimportant; if the good was a medication; or if the data in a study was duplicated in another study. Covidence screening software was used to manage articles for two levels of screening. Data was extracted using an adaption of the Cochrane data collection template. Health outcomes, those that affect the quality or duration of life, are the outcomes of interest. The study was registered with PROSPERO (CRD42017069463). Findings The initial search identified 3370 articles and 59 were included in the final set with a range of 20 to 252 246 participants. The risk of bias assessment revealed that overall, the studies were of medium to high quality. Among the studies included in this review, 80 health outcomes were statistically significant favouring the intervention, 19 health outcomes were statistically significant favouring the control, 141 health outcomes were not significant and significance was unknown for 28 health outcomes. Interpretation The results of this systematic review provide evidence that free goods can improve health outcomes in certain circumstances, although there were important gaps and limitations in the existing literature.


Methods
A search was performed for relevant literature in all languages from 1995-May 2017. Eligible studies were observational and experimental which had at least one tangible item provided for free and had at least one quantitative measure of health. Studies were excluded if the intervention was primarily a service and the free good was relatively unimportant; if the good was a medication; or if the data in a study was duplicated in another study. Covidence screening software was used to manage articles for two levels of screening. Data was extracted using an adaption of the Cochrane data collection template. Health outcomes, those that affect the quality or duration of life, are the outcomes of interest. The study was registered with PROSPERO (CRD42017069463).

Findings
The initial search identified 3370 articles and 59 were included in the final set with a range of 20 to 252 246 participants. The risk of bias assessment revealed that overall, the studies were of medium to high quality. Among the studies included in this review, 80 health outcomes were statistically significant favouring the intervention, 19 health outcomes were statistically significant favouring the control, 141 health outcomes were not significant and significance was unknown for 28 health outcomes. PLOS

Introduction
Disparities in health along socioeconomic lines are well established: groups with lower income and socioeconomic position consistently experience worse health outcomes, including higher rates of mortality. [1,2] One of many possible explanations for better health outcomes among those with higher socioeconomic status is that income allows greater access to tangible goods that can improve health, such as safe shelter, healthy foods, clean water, and essential medicines. Worse health outcomes among lower socioeconomic status groups may be explained by reduced access to education and child care, exposure to hazards such as air pollution or contaminated drinking water, exposure to violence, reduced access to health care services, or discrimination based on gender, ethnicity or other characteristics. [3,4] Some of these potential alternative explanations may be indirectly related to access to tangible goods, such as water filtrations systems that can mitigate effects of contaminated water and medicines that may mitigate the effects of poor access to health care services. The importance of tangible goods has long been recognized through accounting for "non-cash" income, such as the value of housing provided by governments, and by defining poverty based on the cost of tangible goods (as in reference budgets that are baskets of goods and services that are considered necessary to reach an acceptable standard of living for an individual household within a given country, region or city) and essential services rather than based on relative income level. [5,6] If people lack a good that is required for their health and well-being, a simple response is to provide it for free. This approach appears to underpin many governmental and non-governmental programs routinely devote substantial resources to distributing goods to people in need. [7][8][9] Yet it is unclear whether providing goods for free promotes health. Free tangible goods may not be used as intended or at all: their positive health effects may not overcome other causes of poor health, or they may even cause unintended harm (e.g. providing safety equipment such as bicycle helmets could encourage risky behavior). [10] Providing people with free goods could complement other efforts to promote health, such as providing services like healthcare, [11] and providing a Basic Income. [12,13] The receipt of free tangible goods could free up limited household income or resources that would otherwise be consumed in obtaining those goods and this additional disposable income may result in improved health.
We are not aware of any previous systematic effort in the existing scientific literature to assess whether providing free goods promotes health. We systematically reviewed the literature for studies that reported the association between the free provision of tangible goods and health outcomes.

Search strategy
A search strategy was developed in consultation with an information specialist. This systematic review was registered on PROSPERO (CRD42017069463, Aug 30 2017).
We defined "tangible goods" as a physical good or object that could be given to persons or families. We generated a list of items which were hypothesized to be distributed without charge to patients or study participants. The list of items was sent to several other researchers for

Exclusion criteria
We excluded studies in which a service such as advice, health screening procedure or a diagnostic test was provided; if the intervention was primarily a service and the free good was relatively unimportant (e.g. giving participants a voucher for a health service); if the good was a medication (e.g. nicotine replacement, contraception, naloxone kits); or if the data in a study was duplicated in another study (duplicated data was defined as data from the same participant at the same timepoint).

Screening
Covidence screening software [15] was used to manage articles while screening. In level one screening, all titles and abstracts were reviewed to determine if they met the inclusion criteria for the study. Level two consisted of screening the full text of articles to determine whether they met the inclusion criteria. Each article was appraised by two reviewers (LS and HW) for both levels and disagreements were discussed. If the reviewers did not come to a decision, a third investigator (NP) was consulted.
We attempted to include only one report of each health outcome. We excluded reports where both the outcomes and participants were the same as a study that was already included. We included reports where the participants and outcomes only partially overlapped between reports. If multiple reports included the same outcome for the same participants, we included that outcome only once.

Quality appraisal
The quality of each article was appraised by two individual reviewers using the Cochrane Risk of Bias assessment tool for randomized control trials [17] and ROBINS 1 assessment tool for non-randomized control trials. [18] The Cochrane Risk of Bias tool assesses seven potential sources of bias including random sequence generation, allocation concealment, blinding of participants, blinding of outcome assessments, incomplete outcome data, selective reporting, and funding source. [17] The ROBINS 1 tool also assesses seven potential sources of bias including bias due to confounding, bias in selection of participants into the study, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in selection of the reported results. [18] We did not exclude any studies based on the risk of bias assessment.

Presentation of findings
We grouped studies based on the type of free good provided and the outcome reported.

Literature search
The initial search identified 3370 articles of interest. In the first level of screening based on abstract review, 3132 articles were excluded, leaving 238 articles for full manuscript review. This second level of screening removed a further 179 articles yielding a final set of 59 articles which met full eligibility criteria (Fig 1).

Study characteristics
The 59 included studies included a range of 20 to 252 246 participants with a median of 872�5. The length of the studies ranged from two to 180 months with a median of 15�5 months. Of the 59 articles, 29 were randomized controlled trials (RCTs) and 30 were observational studies.
Among the 59 included studies, 45 (76�3%) were from countries that are considered high income according to the 2016 World Bank Report. [20] These countries included the USA (20 studies), Canada (13 studies), United Kingdom (four studies), Norway (two studies), Israel (two studies), Ireland (one study), New Zealand (one study), Australia (one study), and France (one study).Fourteen studies (23�7%) were from countries considered low or medium income by the 2016 World Bank Report. [20] These countries included India (three studies), Cameroon (two studies), and one study each from Mexico, Colombia, Ukraine, Pakistan, Ghana, Kenya, Nigeria, China and Zanzibar.
Among the 59 included studies, the free goods provided were housing (20 studies), food (17 studies), safety equipment (six studies), insecticide treated nets (five studies), hygiene, and water sanitation (six studies) and miscellaneous (five studies).

Risk of bias
Among the RCTs there were: no studies judged to be at a low risk of bias in all domains, one (3�4%) study was at a low or unknown risk of bias for all domains and 28 (96�6%) studies were at a high risk of bias in at least one domain (Fig 2). Among observational studies, there was: one (3�3%) study judged to be at a low risk of bias or no information in all domains, 11 (36�7%) studies at a low or moderate risk of bias or no information for all domains, 13 (43�3%) studies at serious risk of bias in at least one domain (but not at critical risk of bias in any domain), and five (16�7%) studies at critical risk of bias in at least one domain (Fig 3). Risk of bias assessment data is available as S1 Table, Cochrane risk of bias assessment for RCTs and S2  Table, ROBINS 1 risk of bias assessment for observational studies.

Results by type of good
Housing. There were 24 940 participants in the 20 housing studies (there was some overlap in participants between studies; see the Methods section) ( Table 1). All studies were Health outcomes related to the provision of free, tangible goods conducted in either Canada (12 studies) or the USA (eight studies). Nineteen of these studies (95%) had a co-intervention, of which eighteen were "Housing First" programs. For example, in addition to housing, the intervention offered participants treatment for various addictions, mental health challenges and other social supports. [21] The primary reported outcomes in housing studies were stable housing (11 studies, 55%);substance use (10 studies, 50%); psychiatric symptoms or mental health,(eight studies, 40%); quality of life, including QoLI-20, community functioning (MCAS)and community integration (CIS-PHYS and CIS-PSYCH)(eight studies, 40%); health status, including BMI, waist circumference, physical health ailments and health assessments using EQ5D-VAS, and physical SF-12 assessment forms (six studies, 30%); food security (two studies, 10%); and death (one study, 5%). The study durations ranged from six months to 180 months. Housing studies reported a total of 114 outcomes (with duplicates removed), of which 42 were statistically significant, 62 were not significant, and significance was unknown for 10 outcomes. Of the 42statistically significant outcomes, 37 outcomes (from 15 different studies) favoured the intervention, and five outcomes (from two different studies) favoured the control.
Food. There were 307 583 participants in the 17 food studies (Table 2). Food studies were conducted in USA (11 studies), Norway (two studies), Mexico (one study), Colombia (one       study), New Zealand (one study), Ukraine (one study). One study (5�9%)involved a co-intervention consisting of nutrition and education counselling. [41] The most commonly measured health outcome was Body Mass Index (BMI) measured in 12studies (70�6%). The study durations ranged from four to 96 months. Food studies reported a total of 73 outcomes, of which 28 were statistically significant, 41 were not significant, and significance was unknown for four outcomes. Of the 28 statistically significant outcomes, 22 outcomes (from eight different studies) favoured the intervention, and six outcomes (from three different studies) favoured the control group. Hygiene/Water sanitation. There were 10 504 participants in the six hygiene or water sanitation studies (the household was the unit of analysis in two studies) ( Table 3). The free goods distributed were toothbrushes and toothpaste (two studies), a drinking water disinfectant (two studies), and free soap (two studies). The studies were conducted in India (three studies), England (one study), Pakistan (one study), and Israel (one study). Three studies (50%) involved a co-intervention which consisted of social marketing, and educational campaigns. [58][59][60] The most common outcomes were diarrhoea prevalence in three studies Health outcomes related to the provision of free, tangible goods (50%); infection prevalence in two studies (33�3%); and prevalence of dental carries reported in two studies (33�3%). The study durations ranged from nine months to 60 months. These studies reported a total of 34 outcomes, of which 15 were statistically significant, 11 were not significant, and significance was unknown for eight outcomes. All of the 15statistically significant outcomes (from three different studies) favoured the intervention. Insecticide treated nets (ITN). There were 7661 participants in five studies providing ITN ( Table 4). The studies were conducted in Cameroon (two studies), Ghana (one study), Kenya (one study), and Nigeria (one study). Three studies (60%) involved a co-intervention consisting of additional medical care, a social marketing campaign and preventative sulfadoxine-pyrimethamine treatment. [64][65][66] The most common outcomes measured were parasitaemia in three studies (60%); anemia in two studies (33�3%); malaria in two studies (33�3%). Other outcomes included mortality and birth weight. The study durations ranged from four months to 36 months. Eleven outcomes were reported, of which three were statistically significant, and eight were not. Of the three statistically significant outcomes (from three different studies), all favoured the intervention.
Safety equipment. Six studies provided free safety equipment including smoke alarms, hip protectors, mouth guards, and safety equipment for young children (e.g. stair gates and cupboard locks) ( Table 5). We were unable to identify the total number of participants in these studies because some reports did not specify this information. The studies were conducted in England (two studies), USA (one study), Ireland (one study), Israel (one study) and Australia (one study). Five studies (83�3%) involved a co-intervention consisting of educational materials and sessions, [10,[69][70][71] as well as advice, [72] and one study offered stickers to promote the use of safety equipment. [71] The common outcome reported in all six studies was injury. Study duration ranged from six months to 72 months. Safety equipment studies reported a total of 23 outcomes, of which eight were statistically significant, 11 were not significant, and significance was unknown for four outcomes. Of the eight statistically significant outcomes, all eight outcomes (from three different studies) favoured the control and, according to the explanations provided in the articles, this may be been due to infrequent use of the safety equipment. [10,71,73] Miscellaneous. Five studies involved a miscellaneous set of outcomes ( Table 6). The distributed free goods included glucometer test strips for diabetic patients, glucometers, sunscreen, bus passes, and a mobile phone. Three studies (60%) involved a co-intervention consisting of a glucometer (intervention was test strips), [74] educational material and counselling (for the glucometer study) [75] as well as an automated message and calling card to reach participants' primary care physicians (for the mobile phone study) [76]. The outcomes measured included HbA1c, blood glucose, triglycerides, Low Density Lipoprotein (LDL-C), Body Mass Index (BMI), waist circumference, rate of sunburns, and mortality rate. The study durations ranged from two months to 12 months. These studies reported 13 outcomes, of which three were statistically significant, eight were not significant, and significance was unknown for two outcomes. All three statistically significant outcomes (from two different studies) favoured the intervention.

Results by health outcome
In addition to analyzing the results of studies categorized by type of free good distributed to participants, we combined results from the reviewed studies for the health outcomes of mortality and diarrhea because these two outcomes were reported in studies of different categories of goods.
Mortality. Mortality was reported as a health outcome in three studies of mosquito nets (one study), housing vouchers (one study), and mobile phones (one study) including 17 730 participants. The first study gave families with children under five an insecticide treated insect net in Kenya. The study found that receiving a mosquito net was a significant predictor of reduced mortality (rate ratio: 0�56; 95% confidence interval (CI): 0�33-0�96). [65] The second study gave a housing voucher to families of children living in public housing in the USA. [24] Receiving a housing voucher was not a significant predictor of mortality in any of the 3 Health outcomes related to the provision of free, tangible goods categories; deaths from disease (p = 0�84), deaths by homicide (p = 0.81),and accidental deaths (p = 0�19). [24]The final study gave phones to pregnant women in Zanzibar. [76] Mortality was recorded in three ways: stillbirth (unadjusted odds ratio (UOR): 0�62; 95%CI: 0�31-1�22), perinatal mortality (UOR: 0�49; 95%CI: 0�27-0�90), and neonatal mortality (UOR: 0�85; 95%CI: 0�37-1�95). Receiving a free phone significantly reduced perinatal mortality. [76] Diarrhea. Diarrhea was reported as a health outcome in four studies of food (one study), and hygiene and water sanitation (three studies), which included 8382 participants. The first study conducted in Pakistan included households in squatter settlements receiving either bleach, hand washing supplies, flocculant-disinfectant, or flocculant-disinfectant plus hand washing. [61] The authors concluded that receiving any of the free goods, as well as the intense community-based intervention, which included meetings and presentations to community leaders and residents about the importance of hygiene wand water contamination, reduced the daily longitudinal prevalence of diarrhoea; however, the level of statistical significance was not reported. [61]The second study, conducted in Colombia, gave primary school children a school snack. [47] The authors found that the rate of days per child year of diarrhoea (unadjusted rate ratio (URR):0�68; CI: 0�63-0�73), and diarrhoea with vomiting (URR: 0�63; CI: 0�52-0�75) were significantly reduced with the provision of a school snack. [47] The third study, conducted in India, gave children under the age of five sodium dichloroisocyanurate tablets. [59] The authors found that the longitudinal prevalence of diarrhoea for children given sodium dichloroisocyanurate tablets was not significantly different from the control (prevalence ratio: 0�95; CI: 0�79-1�13). [59]The final study, conducted in India, distributed soap to households with children under five, and outcomes were assessed for the target children, as well as their family, including siblings. [60] The authors reported significant relative risk reductions (RRR) in diarrhoea prevalence related to the provision of free soap among four groups: target children (RRR: 25�3%; CI 36�6-2�3); children aged five and under (non-target) (RRR: 32�5%; CI 41�1-3�8); children aged six-15 (non-target) (RRR: 30%; CI 38�7-6�6); and whole families (observed RRR 30�7%: CI 37�5-5�5). [60] As such, three of the four studies reported that diarrhoea was significantly reduced with the provision of free goods.

Interpretation
The results of this systematic review provide evidence that free goods can improve health outcomes in certain circumstances, although there are also important gaps and limitations in the  Health outcomes related to the provision of free, tangible goods participants (most were individuals but some were households) that examined the health effects of free goods, the most commonly studied free goods were housing (20 studies) and food (17 studies). Among the 268 total outcomes reported, the most commonly reported outcomes were housing retention in 12 housing studies and BMI in 12 food studies. Four RCTs were deemed to be unclear or at high risk of bias, and one non-RCT was rated as serious, critical or no information, in all risk of bias categories. Therefore, overall the studies were of medium to high quality in terms of bias. Among the studies included in this review, 80 health outcomes were statistically significant favouring the intervention, 19 health outcomes were statistically significant favouring the control, 141 health outcomes were not significant, and significance was unknown for 28 health outcomes. The rationale underpinning how the provision of free tangible goods impacts health was typically not stated in the reviewed studies. However, we identify four related concepts that help us understand the rationale for providing free tangible goods. First, facilitating access to a good that is capable of promoting health should promote health unless there are unintended negative effects or implementation problems. We did in fact find some studies where those receiving a free good had worse health outcomes (e.g. hip protectors were associated with an increased risk of hip fractures). [71] Second, if poverty is defined, at least partially, as being unable to afford tangible goods (and services) in a market-based economy, [79] then studies examining the impact of free good provision on health describe the effect of poverty reduction on health. Findings from these studies could then be considered alongside studies of other interventions aimed at reducing poverty, such as a basic income as a complementary approach to reducing poverty. [12,13] Third, the free provision of goods could be understood as "noncash" income that is valued similar to its cash equivalent after being appropriately discounted. [6] Fourth, having certain tangible goods can be understood as fulfilling a basic human right (e.g. the right to adequate housing, the right to adequate nutrition and clean water). [80] The provision of such goods could be seen as achieving social justice and could have positive impacts not only for individuals but also for their communities.

Comparison with prior studies
To the best of our knowledge this is the first systematic review to examine a wide range of free tangible goods and their effects on health. One recent systematic review and narrative analysis of 31 Housing First studies found mixed results for the impact of providing free housing for substance abuse and psychiatric symptoms, a clear benefit for housing stability, and a benefit for quality of life. These findings generally align well with ours. [81] A number of studies have examined whether people who were given free goods use them or resell them. One such study conducted among pregnant women and households with young children in Uganda, for example, investigated this concept with the provision of free long-lasting insecticide treated mosquito nets. [82] This study assessed the willingness to pay for a mosquito net and willingness to sell a mosquito net given for free by simulating market exchanges. Seventy-three percent of people who received free nets were unwilling to accept the maximum price offered to part with even one of their nets. [82] Most people who were given free nets were not likely to resell their nets and in fact did use them for their intended purpose. [82] Other studies have investigated using financial investments to complement health interventions and further improve health outcomes. A non-randomized controlled assessment from sub-Saharan Africa, in which simultaneous investments were made in agriculture, the environment, business development, education, infrastructure, and health in rural village sites with high baseline levels of poverty and under nutrition, found that mortality rates in young children decreased by 22% in study sites relative to baseline. [83] Reductions in poverty, food insecurity, stunting, and malaria parasitemia were also reported in study sites. [83]

Strengths and limitations of our study
Due to the great variety of free goods with potential to impact health, the design of a search strategy was challenging and we may have inadvertently omitted some key search terms. The wide array of interventions and outcomes meant that we could not perform a meta-analysis of results. The broad approach allowed us to include an interesting array of studies of different free tangible goods. Some studies involved co-interventions (e.g. almost all housing studies involved other supports in addition to free housing) and this limits the ability to determine whether the free good or the co-intervention affected health outcomes. We also excluded many studies that provided free tangible goods, including clean needles, condoms, and baby cribs, but did not report a health outcome. The literature may be biased towards studies of items with a less certain benefits. In other words, researchers may have decided not to study certain goods which are very likely to be beneficial (e.g. condoms, clean needles) and some such studies may not be ethical (i.e. it may be difficult to study the free provision of an item that is very likely to be beneficial). Some of the Housing First studies were overlapping as different reports included some of the same participants and some of the same outcomes, so we attempted to strike a balance between not excluding results and not counting the same results twice.

Conclusions and future work
Findings of this systematic review suggest that providing free tangible goods can promote health in certain circumstances. Additional high-quality studies of different goods are needed. Future work should also focus on the contexts in which free goods are most beneficial and explicitly state the theory or theories underpinning each study or intervention.