Basic needs, stress and the effects of tailored health communication in vulnerable populations

Abstract

This study examined whether unmet basic needs (food, housing, personal and neighborhood safety, money for necessities) and perceived stress affect recall of and response to a tailored print intervention one month later. Participants (N = 372) were adults who had called 2-1-1 Missouri between June 2010 and June 2012. A series of path analyses using Mplus were conducted to explore the relationships among basic needs, perceived stress, number of health referrals received in a tailored intervention, recalling the intervention and contacting a health referral. Participants were mainly women (85%) and African-American (59%) with a mean age of 42.2 years (SD = 13.3; range 19–86); 41% had annual household income <$10 000. Unmet basic needs were positively associated with increased levels of perceived stress, which, in turn, were negatively associated with recalling the intervention and calling any of the health referrals provided. Tailored printed interventions may be less effective in populations with acute unmet basic needs. More broadly, the effectiveness of minimal contact behavioral interventions might be enhanced by simultaneous efforts to address unmet basic needs.

Introduction

Meta-analyses and narrative reviews have concluded that tailored print materials are more effective than non-tailored ones in changing health behavior [1–8]; they have also demonstrated that study samples in tailoring research have been predominantly white and educated [1, 2]. In contrast, multiple tailoring studies among underserved populations—including those with low-income, minorities and Medicare patients—have shown that tailoring is not more effective than comparison conditions [9–13]. For instance, in a study of 1574 low-income and minority women overdue for breast or cervical cancer screening, tailored letters were no more effective than generic form letters in increasing mammography or Pap testing in the next 12 months [12].

What might explain these apparently differential effects of tailoring? Previous studies have called for new research to identify reasons tailoring works in some groups and not others [10, 11], but stopped short of suggesting specific variables. We propose that the challenges of living in relative deprivation may interfere with critical processes necessary for tailoring effects. Specifically, that unmet basic needs and perceived stress compromise exposure, attention and information processing, and therefore recall and response to tailored health communication.

Basic needs like food, shelter, safety and money for necessities sometimes go unmet in economically disadvantaged populations. Having unmet basic needs predicts adverse physical and mental health outcomes and mortality [14, 15] and is a strong predictor of psychological stress [16], which also adversely affects health [17, 18] and is experienced disproportionately by low-income and minority populations [19].

People living in poverty often have such overwhelming needs in other areas of life that disease prevention is perceived as irrelevant, or, at best, secondary to these priorities [20]. Mullainathan and Shafir [21] suggest that this is due to effects of ‘scarcity’ on the brain. Having unmet basic needs, they argue, orients the mind towards filling those needs and reduces one’s available attention and cognitive bandwidth that can be devoted to other concerns.

Building upon this explanation, we hypothesize that unmet basic needs and perceived stress will reduce the likelihood that individuals pay attention to and act on tailored health referrals. Using path analysis on data from a recently completed randomized trial [22] we examine the extent to which four unmet basic needs (insufficient food and income, inadequate housing and unsafe neighborhoods and personal safety) and perceived life stress predict recall of a tailored intervention and contacting a recommended health referral contained in the intervention one month later.

Method

Data analysed for this study were collected as a part of a randomized trial among callers to the 2-1-1 helpline, an information and referral service accessed by dialing the three digits 2-1-1, similar to 9-1-1 for emergencies or 4-1-1 for directory assistance. 2-1-1 serves predominantly low-income individuals and families by linking them to needed health and social services in their community. Studies have shown that the health needs of 2-1-1 callers greatly exceed those of the general population [22–25]. The trial compared the effects of verbal health referrals from 2-1-1 information specialists with and without tailored print reminders and telephone health coaches (i.e. navigators) [22].

Participants were adults who had called 2-1-1 Missouri between June 2010 and June 2012. After receiving standard service from 2-1-1, a random sample of callers completed a brief cancer risk assessment to identify their need for breast, cervical and colorectal cancer screening, HPV vaccination, smoking cessation, and having a smoke free home policy. Callers that had at least 1 cancer control need were invited to participate in the study.

Across all study conditions, referrals for needed cancer control services were made to health care providers, organizations and/or agencies that were near the caller’s home and provided free or low-cost services. Referrals were provided for up to 3 cancer control needs. Baseline assessments were completed at the time of the 2-1-1 call and telephone follow-up interviews were conducted 1-month later.

Analyses in this article included only participants in the tailored reminder group (n = 510) because other participants did not receive printed referrals by mail. The tailored materials were mailed to participants within one working day of their receiving the verbal referral. The tailored reminder was a 4-page full-color booklet containing a brief story about a similar (hypothetical) peer, information the participant would need to contact each cancer control referral he or she received, and information describing why each of the recommended cancer control service was important. Information and photos in the reminder were tailored based on each caller’s age, sex, race/ethnicity, children in the home and reason for calling 2-1-1. About three quarters of participants in the tailored intervention condition completed 1-month follow-up (n = 373; 73%) and constitute the final sample for analyses.

Measures

Participant characteristics, including demographics, basic needs and perceived stress were assessed at baseline. The outcomes of interest—recalling the intervention and calling a health referral—were assessed at 1-month follow-up. Both sets of questions were administered by telephone interview.

Basic needs

We adapted 7 items from Segal’s [26] Personal Empowerment scale and a second scale developed by Blazer et al. [27]. Items assessed the perceived likelihood that one’s housing, food, safety and financial needs would be met in the coming month. Five questions asked, ‘How likely is it that <basic need> in the next month?’ Basic needs included ‘you will get enough to eat’, ‘you will have a place to stay’, ‘someone will threaten to hurt you physically’, ‘you will have enough money for necessities like food, shelter and clothing’ and ‘you will have enough money to deal with unexpected expenses?’ Responses were on a 4-point scale (1 = very unlikely to 4 = very likely).

Other items asked ‘Considering the number of people living in your home, would you say you have…’ (1 = not enough space, 2 = about the right amount of space, 3 = more than enough space); and ‘How would you rate the safety of your neighborhood?’ (1 = very unsafe, 4 = very safe). From these, we created an index variable of unmet basic needs. If a need was very unlikely or unlikely to be met, it was considered unmet. The number of unmet basic needs was then summed, yielding a value from 0 (i.e. no unmet basic needs) to 7 unmet basic needs.

Perceived stress

We administered the 4-item Perceived Stress Scale (PSS; [28, 29]), a short version of the PSS that is a useful tool when data are collected by phone. All items begin with the stem ‘In the last month, how often have you felt…’ and conclude with ‘that you were unable to control the important things in your life?’; ‘confident about your ability to handle your personal problems?’; ‘that things were going your way?’; and ‘difficulties were piling up so high that you could not overcome them?’ Responses were on a 5-point scale (1 = very often, 2 = fairly often, 3 = sometimes, 4 = almost never, 5 = never). We reverse coded items when necessary so that higher values were associated with more perceived stress. Per Cohen (1983), we summed the responses to these four items as a measure of perceived stress such that 4 = the lowest level of perceived stress and 16 = the highest level of perceived stress [28].

Recall of tailored materials

We assessed unprompted recall of the tailored intervention by asking ‘Did you get information in the mail from My 2-1-1?’ (yes/no/don’t remember). If a participant answered ‘no’ or ‘don’t remember’, we assessed prompted recall by saying, ‘The information would have come in a big blue envelope that said ‘My 2-1-1’. Inside the envelope was something that looked like a little magazine and had information about health resources in your community. Do you remember getting this?’ (yes/no). We considered a participant to have recalled the intervention if he or she answered ‘yes’ to either the unprompted or prompted recall questions.

Recall of verbal health referral

We assessed unprompted recall of receiving verbal health referrals from a 2-1-1 information specialist by asking, ‘In addition to 2-1-1 referrals, some also received health referrals. Did you receive any health referrals?’ (yes/no/don’t remember). If a participant answered ‘no’ or ‘don’t remember’, we assessed prompted recall by saying, ‘The referrals would have been about things that you can do to help improve your health. Do you remember getting any referrals?’ (yes/no). We considered a participant to have recalled a verbal referral if he or she answered ‘yes’ to either the unprompted or prompted recall questions.

Called any health referral

For each health referral a participant received, we asked at 1-month follow-up if they had contacted that referral. From these responses we created a dichotomous variable indicating whether a participant had contacted any or none of the referrals.

Data analysis

We conducted a path analysis using Mplus 7.11 to determine the relationships among basic needs, perceived stress, number of health referrals received (obtained from study records), recalling tailored materials and calling a health referral. This method was chosen in order to examine the relationships between multiple variables simultaneously.

A comparison model also was tested to assess the possibility that verbal health referrals, not printed tailored materials, had prompted participants to contact a health referral. This model determined the relationships among basic needs, perceived stress, number of health referrals received, recalling verbal health referrals and calling a health referral.

Because the dependent variables in both models were dichotomous, probit models by weighted least squares mean variance (WLSMV) were estimated. Age was included as a covariate in both models because participants are eligible for certain cancer control services only after they reach a certain age (e.g. breast and colorectal cancer screening). We also examined the indirect effect of basic needs on recalling the tailored intervention and on calling a referral.

Model fit was assessed using the Tucker-Lewis index (TLI) and comparative fit index (CFI). Fit indices >0.95 indicate adequate fit to the data. We also report the Root Mean Square Error of Approximation (RMSEA). Low values (<0.06) indicate adequate model fit [30].

Results

Participants

Participants who were not reached at 1-month follow-up (n = 127; 27%) did not differ significantly from respondents with respect to age, gender, number of unmet basic needs, perceived stress or number of health referrals received.

The final sample used in all analyses (n = 372) comprised mostly women (85%) who were African-American (59%) or white (31%) and reported annual household income <$10 000 (41%). Their average age was 42.2 years (SD = 13.3; range 19–86) and 60% reported having 12 or fewer years of education (Table I).

Table I.

Participant characteristics (N = 372)

Mean age (years; SD)	42.2 (13.3)
Gender
Female	84.7%
Race/ethncity
Black	59.0%
White	31.3%
Other	9.7%
Income
≤$10 000	42.2%
Education
≤12 years of education	60.0%
Number of health referrals
1	55.4%
2	32.0%
3	12.6%
Unmet basic needs
0	5.4%
1	16.7%
2	30.1%
3	31.5%
4	11.3%
5	4%
6	1.1%
7	0%
Perceived stress (mean; SD)	8.76 (3.5)
Recalled health referrals	67%
Recalled tailored intervention	57%
Called any health referral	22%

Basic needs, perceived stress and health referrals

Callers had a mean of 2.4 unmet basic needs (SD = 1.2; range 0–7). Half of the sample (55%) received one health referral, 32% received two and 13% received three health referrals. The mean score for perceived stress was 8.81 (SD = 3.56; range 0–16).

Recall of tailored intervention and contacting health referrals

At 1-month follow-up, 67% of participants reported that they remembered receiving verbal health referrals, 57% of participants reported that they remembered receiving information in mail from 2-1-1 and 22% reported contacting at least one of the cancer control referrals they received (Table I).

Path analysis

Figure 1 illustrates the model postulating direct paths between (i) unmet basic needs and perceived stress; (ii) unmet basic needs and number of health referrals received; (iii) perceived stress and recall of the tailored intervention; (iv) number of health referrals received and recall of the tailored intervention; and (v) recall of the tailored intervention and contacting any of the health referrals received. The model also postulated indirect effects of perceived stress and number of health referrals received on recall of the tailored intervention and contacting any of the health referrals received.

Fig. 1.

Direct and indirect paths between study variables.

Fit indices indicated a good fit of the model to the data (TLI = 1.1; RMSEA = 0.000 with 90% confidence interval = 0.000 to 0.120; CFI = 1.0). Figure 1 shows the final path model with standardized coefficients. Unmet basic needs was positively and significantly correlated with number of health referrals received and perceived stress, which in turn were both significantly correlated with recall of the tailored intervention and contacting any of the health referrals received, but in different ways. Perceived stress was negatively related to recall of the tailored intervention, while number of health referrals was positively related to contacting any health referrals.

Recall of the tailored intervention was positively related to contacting any of the health referrals received. Results also confirmed the indirect path between unmet basic needs and any recall (P = 0.036; mediating variable: perceived stress).

No other significant paths were found.

Comparison model

This model postulated direct and indirect paths identical to those in the original model, except that recall of verbal referral replaced recall of tailoring interventions. Fit indices indicated a good fit of the model to the data (TLI = 0.797; RMSEA = 0.046 with 90% confidence interval = 0.000–0.094; CFI = 0.927), but the path from recalling verbal referral to contacting a referral was not significant (P > 0.5).

Discussion

As a participants’ number of unmet basic needs increased, so did their level of perceived stress, which predicted lower rates of recall of a tailored intervention one month later. Recall was in turn associated with higher rates of acting on the recommendations in the tailored intervention by contacting a health referral for a needed cancer control service. These findings suggest a plausible pathway through which unmet basic needs in vulnerable populations may interfere with the effects of tailored health behavior interventions. Confidence in these findings is bolstered by the fact that recalling a verbal health referral did not produce the same results.

While our analyses demonstrate a link between basic needs, perceived stress and responses to tailoring, they do not identify specific mechanisms or processes through which this occurs in vulnerable populations. For example, participants experiencing economic hardship may not open unexpected mail from unknown sources for fear it contains a bill or warning. Their mailing address might be a temporary residence with friends or family members who don’t pass along the mail.

Alternatively, findings may reflect participants’ inability to attend to the intervention in the face of higher priority life demands. Mullainathan and Shafir [21] and others demonstrate that a tunnel-vision focus on unmet needs can lead to short-term thinking, a focus on immediate stressors, poor decision-making, diminished behavioral control and less planning for prevention [31–33].

In the context of tailoring, this could influence the extent to which individuals pay attention to an intervention. There is some support for this explanation. In a 2003 study comparing three types of tailored materials, participants were assessed at baseline for ‘present time orientation’—the tendency to focus on immediate or short-term consequences in one’s decision-making. At 6-month follow-up, across all study groups, participants who were classified as having a present time orientation were significantly less likely to have read and remembered the tailored materials than those who did not have a present time orientation [34].

Studies have shown that the advantage of tailored over non-tailored communication lies in its ability to elicit greater and more central message processing while also raising recipients’ expectations that they are receiving something made just for them [35–37]. But these advantages are lost if people never attend to tailoring in the first place, and unmet basic needs appear to have disruptive effects on intervention exposure, recall and attention.

While the focus of this article was on tailored health communication interventions, it seems likely that unmet basic needs could interfere with other types of health behavior interventions, especially communication-based approaches. Future research should test this assumption empirically. Nonetheless, the focus on tailoring here is justified by its status as an evidence-based intervention and its presence in evidence compendiums and practice recommendations [38]. This is not the case for many other mailed print interventions and therefore heightens the relevance of potentially contradictory information.

Also untested but worth exploring is whether the potential effects of unmet basic needs are unique to mailed interventions or apply more broadly, including to e-health and m-health modalities being advanced as promising solutions for reaching economically vulnerable populations [39]. Low health literacy might also contribute to the effects we observed either through avoidance behavior (e.g. not opening printed information) or inability to read, process and learn from the information. Both potential explanations lend themselves to empirical study.

The analyses conducted for this article are unique in their consideration of basic needs and perceived stress as factors influencing tailoring effectiveness, and in their focus on a low-income and racially diverse study sample. But many more theoretical and empirical insights are needed to advance our understanding, including improving upon the limitations of this study.

For example, because basic needs and perceived stress were both measured at the same time, we know they are associated but cannot say with certainty that one caused the other. In addition, while our measures of basic needs (7 items) and perceived stress (4 items) are relative strengths of the study, our outcome variables of recall and response were comparatively simple. These limitations can be explained in large part as consequences of a post-hoc secondary analysis drawing on available data. Future research that is designed specifically to explore these relationships will easily improve upon our article’s shortcomings.

Finally, it is possible that our basic needs variables are only a proxy for some other indicator of deprivation. Although we measured education and income among study participants and found no relationships between these socio-demographic variables and our outcomes, it is possible that some unmeasured variable might explain the study’s findings. However, newly published research lends support to our basic needs hypothesis. In a larger, longitudinal study of 2-1-1 callers, Thompson and colleagues found that those whose basic needs problems were resolved one month after calling were more likely to act on a health referral at 4-month follow-up than those whose problems were not resolved at 1-month follow-up [40]. Because these health behavior actions occurred over such a short interval, it is unlikely they could have been caused by changes in other indicators of deprivation, like years of education.

It would be premature to base practice recommendations on the findings from one set of analyses. In the specific context of tailoring, mailed printed interventions appear to have limitations for populations with acute unmet basic needs. Effectiveness of tailored print materials might be strengthened by supplementing them with social workers or navigators who can help address participants’ unmet basic needs. The effectiveness of these combined strategies for specific population subgroups should be studied further.

Helping people meet their basic needs may have positive spillover effects that can increase their capacity or ‘bandwidth’ to focus on longer-term goals, including health improvement [30]. This is an appealing, if intuitive, proposition for those working to eliminate health disparities, and one that is largely untested as it relates to the effects of health behavior interventions. We view it as a great opportunity for research that could have meaningful social impact.

Funding

This study was supported by funding from the National Cancer Institute (P50-CA095815).

Conflict of interest statement

None declared.