Discussion

In a study of 769 low-income adult smokers visiting a single ED in the north-eastern US, subjects were easiest to reach by phone in the first week of the month, with decreased odds of contact in subsequent weeks, and lowest in the final week. The association between timing of call and odds of contact, to our knowledge has not been previously reported and suggests a potential method to reduce attrition in a broad range of research involving low-income populations.

Although likely multifactorial, one plausible hypothesis for our decreased ability to reach participants during the last week of the month may be related to decreased access to active telephone service. Based on spending patterns observed in low-income households (Stephens, Reference Stephens2003) and literature describing increased food scarcity at the end of the month, we suspect that the variation in ability to reach low-income patients is most likely related to greater availability of active telephone service, particularly cell phones, early in the month.

In the United States, virtually all public assistance including supplemental security income (SSI), SSI with disability insurance, Temporary Assistance for Needy Families, and Supplemental Nutrition Assistance Program are distributed in the first few days of the month (“DSS: How to Access Cash and Supplemental Nutrition Assistance Program (SNAP) Benefits,” n.d.; Security Administration & of Communications, 2012). As the month progresses, available resources decrease and food insecurity increases in low-income populations (Tarasuk, McIntyre, & Li, Reference Twitchell, Hertzof, Klein and Schuckit2007). We suspect, therefore, that more participants are able to pay for their phone service early in the month, although we acknowledge the existence of other possibilities such as increased time away from home at work or procuring food or decreased time available to answer the phone.

Prior studies demonstrated the improved financial position of low-income households early in the month. These households demonstrate increased spending early in the month, after monthly paycheck or public assistance disbursement, and increased food insecurity towards the end the month (Hamelin, Beaudry, & Habicht, Reference Hamelin, Beaudry and Habicht2002; Hastings & Washington, Reference Hastings and Washington2010; Seligman, Bolger, Guzman, López, & Bibbins-Domingo, Reference Seligman, Bolger, Guzman, López and Bibbins-Domingo2014; Stephens, Reference Stephens2003; Tarasuk et al., Reference Tarasuk, McIntyre and Li2007; Thompson et al., Reference Thompson, Taren, Andersen, Casella, Lambert and Campbell1988). An analysis published in 2003 describes a 7–20% increase above mean expenditures in the week following receipt of Social Security benefits by families whose benefits constitute more than 70% of their income, with an 80% increase in spending on the day of benefit receipt (Stephens, Reference Stephens2003). Correspondingly, multiple studies report increased food insecurity in low-income populations with declining food expenditures, decreased consumption of vegetables and milk products, decreased overall calorie consumption, and increased use of soup kitchens at the end of the month, all of which were attributed to exhaustion of monetary resources (Hamelin et al., Reference Hamelin, Beaudry and Habicht2002; Hastings & Washington, Reference Hastings and Washington2010; Tarasuk et al., Reference Tarasuk, McIntyre and Li2007; Thompson et al., Reference Thompson, Taren, Andersen, Casella, Lambert and Campbell1988). Thompson et al. (1988) reported a 43% increase in the number of meals served by an Upstate New York soup kitchen in the last compared to the first week of the month (Thompson et al., Reference Thompson, Taren, Andersen, Casella, Lambert and Campbell1988).

This phenomenon is consistent with California hospital admission data from 2000 to 2008. A study evaluating hospital admissions found a 27% increase in admissions for hypoglycaemia for low-income diabetics in the last calendar week of the month compared to the first week (Seligman et al., Reference Seligman, Bolger, Guzman, López and Bibbins-Domingo2014). A more recent study, found increased odds of ED visit or hospital admission for hypoglycaemia in patients with household incomes less than the national median during Week 4 compared to earlier weeks (Basu, Berkowitz, & Seligman, Reference Basu, Berkowitz and Seligman2017). Overall, studies show that low-income households struggle with paying for basic needs towards the end of the month, including food, which would be a heavily competing priority for households who purchase cell phone minutes only when they can afford them.

Limited telephone service at the end of the month due to competition for finite resources would be much more likely in households with predominantly cellular phone service plans, in which one buys minutes as needed rather than a monthly bill for a land line. Data from the National Health Interview Survey (NHIS) from 2010 to 2012 showed that the proportion of wireless-only phone services in households below the poverty line, at 100–200% of the poverty line, and >200% of the poverty line, increased, respectively, from 42.8–54.3%, 35.2–45.9%, and 21.7–33.2% (Blumberg & Luke, Reference Blumberg and Luke2013) ²⁰. Another recent study reported cell phone ownership of 74% among 1,898 individuals with annual incomes less than $20,000 (Purnell, Griffith, Eddens, & Kreuter, Reference Purnell, Griffith, Eddens and Kreuter2014).

Disconnection of phone service occurs for many low-income patients. In a 2012 survey of predominantly low-income (70% with annual income < $35,000) participants, 27.3% reported disconnection of phone service in the previous 24 months (Warner et al., Reference Warner, Glasgow, Emmons, Bennett, Askew and Rosner2013). Another study of repeat offenders for DUI reported disconnected telephone service in one-third of participants. Over time, investigators were able to complete follow-up in 93 of the 243 (39.7%) participants that had disconnected phone service (Kleschinsky et al., Reference Kleschinsky, Bosworth, Nelson, Walsh and Shaffer2009). This suggests that telephone disconnection is frequently a transient phenomenon, although the study did not examine the association between the timing of calls and the likelihood of contact. It is also possible that the variation is not entirely due to disconnected phone service, as cell phone users may engage in more aggressive screening of phone calls later in the month to “save” minutes for emergencies. Although they did not evaluate whether time of month was related, study participant concern over the personal cost of using cell phone minutes to participate in research follow-up has been previously described as a reason for refusing to participate in follow-up phone calls (Kleschinsky et al., Reference Kleschinsky, Bosworth, Nelson, Walsh and Shaffer2009). As supported by NHIS data (Blumberg & Luke, Reference Blumberg and Luke2013) that suggest increase rates of wireless-only households in low-income populations, we suspect that most of our participants were cell phone users, although unstable housing likely played some role in our contact patterns.

Though disconnection or non-use of telephone services as economic resource scarcity increases during at the end of the month is our leading hypothesis, other factors are plausible. Participants may have been more difficult to reach at the end of the month if they provided a land line phone for contact, if they were more frequently eating outside of the house in the homes of friends or family with less food insecurity, or if they were more frequently eating out of the house at soup kitchens, which have been shown to be busier at the end of the month (Thompson et al., Reference Thompson, Taren, Andersen, Casella, Lambert and Campbell1988). Another possibility is that individuals with tip-based or cash-based employment may be working more, or have “picked up extra shifts” at the end of the month in an attempt to offset the monthly fluctuation of resources and are therefore less available to participate in study follow-up.

Regardless of the aetiology of our findings, this work may have implications both for clinical care and research. Lessons learned from this study include (1) the importance of persistence – more follow-up calls continue to increase the likelihood of reaching an individual, (2) variation in time of day, day of week, and time of month may increase likelihood of reaching a patient for follow-up, (3) texting was an effective method for reaching participants in some cases (although we did not consistently collect text data, we were able to establish contact and arrange a convenient time for telephone follow-up in some cases).

The study has several limitations. We did not ask subjects what kind of phone service they have, or whether they “conserve” minutes as the month progresses, so we cannot make inferences about our inability to reach specific subjects. We also do not have information about the billing cycle for participants’ telephone service plans, although anecdotal evidence from our subjects indicates that many are on “pay as you go” plans and purchase additional cell phone minutes when needed, and when they have funds to do so. As a proxy for low-income, we used insurance status. Most subjects on Medicaid or who are uninsured are low income, but we did not explicitly ask about household income. Finally, we did not ask our subjects whether they receive public assistance (other than Medicaid), or the relationship between receipt of public assistance and their household income spending habits.