A Facebook Follow-Up Strategy for Rural Drug-Using Women

Abstract

Purpose

Facebook (FB) use has grown exponentially over the past decade, including in rural areas. Despite its popularity, FB has been underutilized as a research follow-up approach to maintain contact with research participants and may have advantages in less densely populated areas and among more hard-to-reach, at-risk groups. The overall goal of this study was to examine FB as a supplemental follow-up approach to other follow-up strategies with rural drug-using women.

Methods

Face-to-face interviews were conducted with randomly selected women who completed baseline interviews in 3 rural jails in 1 state. Analyses focus on participants who were released from jail and were eligible for 3-month follow-up (n=284). Bivariate analyses were used to examine differences between FB users and non-users, and multivariate logistic regression models examined predictors of 3-month follow-up participation and being located for follow-up using FB.

Findings

About two-thirds (64.4%) of participants were regular FB users. Bivariate analyses indicated that FB users were younger, more educated, and more likely to have used alcohol in the 30 days before incarceration but less likely to have a chronic health problem. Regression analyses indicated that rural FB users had more than 5 times the odds of being located for the 3-month follow-up interview, even after controlling for other variables. There were no significant predictors of being followed up using FB.

Conclusions

Findings suggest that FB is widely used and well accepted among rural drug-using women. Among hard-to-reach populations, including those in rural, geographically isolated regions, Facebook serves as a method to improve participant follow-up.

Social media use has grown exponentially in recent years. Facebook (FB) is the most used social networking site with billions of users worldwide, including two-thirds of American adults.^1,2 The accessibility and portability of FB makes it ideal for communication and sustained connectivity across relationships,^3,4 particularly in areas with geographic isolation such as rural communities.⁵ As costs of smartphones and other devices decrease, FB will likely reach previously untapped groups. In 2011, 83.7% of US households reported cell phone ownership and 78.5% reported computer ownership.^6,7

For isolated populations, social media serves as a social outlet, a tool for maintaining relationships, and a means for connecting with the outside world.^8-10 Specifically, Gilbert and colleagues¹¹ found that women in rural areas represent a significantly larger proportion of social networking site users compared to urban areas. Similarly, the Internet could serve as a method for building social support for rural women experiencing limited mobility.¹²

In social science research, FB has been used for study recruitment,^13-15 as a prevention, screening, and survey tool,^16,17 and to deliver interventions and promote health.^18-20 However, the feasibility and effectiveness of social media for fostering outcomes and decreasing attrition in health research has not been adequately examined.²¹ Despite the use of FB as a research tool, it has been underutilized to maintain contact with study participants.²² Problems encountered in longitudinal studies such as attrition, and the difficulties in conducting research with hard-to-reach populations, may make FB an appealing low-cost resource for tracking and communicating with study participants.^23,24 FB may also be a more appropriate follow-up method among transient research populations including those with substance use and mental health problems, and individuals who are criminally involved.^25,26

While studies indicate that FB is a well-supported research tool, studies examining FB use among rural populations are limited, including FB as a follow-up approach. The purpose of this study is to compare FB users and nonusers in a sample of rural drug-using women transitioning to the community from jail and to examine FB as a method to enhance study follow-up. Specifically, this study examines the relationship between FB use and follow-up completion and explores predictors of being located using FB compared to other follow-up approaches, including demographics, substance use, criminal, and physical and mental health histories.

Methods

Participants

As part of a NIDA-funded, IRB-approved study, face-to-face interviews were conducted between November 2012 and September 2015with randomly selected female participants recruited from 3 jails in the rural Appalachian area of 1 state. The jails were located in rural counties with Beale Codes of 7 and 9, classifying them as non-metropolitan counties not adjacent to a metropolitan area.²⁷ Each female had an equal chance of being selected if she had a projected release date between 2 weeks and 3 months (verified by online jail records). Participants were randomly selected for screening using the Research Randomizer computer-based program (www.randomizer.org).

At the time of this study, 400 participants had completed the baseline interview; 304 of whom were released from jail and eligible for 3-month follow-up. Of those eligible, 20 had missing data on at least one variable of interest and were omitted from analyses, resulting in a final sample of 284.

Procedures

Randomly selected women were invited to complete a short screener that included the NIDA-modified Alcohol, Smoking and Substance Involvement Screening Test (NM-ASSIST), and a risky sexual behavior screener. Study eligibility has been defined elsewhere and is summarized by: 1) NIDA-modified ASSIST (NM-ASSIST) score of 4+ for any drug, indicating at least moderate risk for substance abuse²⁸; and 2) engagement in at least 1 sex risk behavior in the 3 months before incarceration.²⁹

Baseline interviews were conducted in a private room in a jail. Trained rural female interviewers asked participants about substance use, mental health, and criminal histories using laptops with Computer Assisted Personal Interview (CAPI) software. Participants were paid $25 for their time. Detailed follow-up tracking information was collected at baseline. To schedule follow-ups, project staff first attempted to contact participants using FB (if participant indicated they used FB at baseline), followed by telephone, mail, and lastly, a home visit.

Measures

Demographics

Baseline demographic information included age, race/ethnicity (1=white, 0=non-white), years of education, relationship status (1=in a relationship; 0=single), employment in the 6 months prior to jail (1=employed at least part time; 0=unemployed), income, and driver's license status (1=currently valid license, 0= no valid license).

FB Use

FB information was collected at the baseline interview and for this study, FB users are those participants who were active on FB during the follow-up period (ie, after release from jail, viewed, and/or responded to messages sent from the data coordinator through the confidential, invite-only FB study site; 1 = FB user, 0 = nonuser).

Substance Use, Criminal, and Health Histories

Participants were asked at baseline about 1) substance use patterns and injection behaviors during the 30 days before incarceration (1=yes, 0=no); 2) age at first arrest, the number of arrests and the number of incarcerations, and whether they had been in prison (1=yes, 0=no); 3) experienced symptoms related to anxiety, depression, and PTSD during the past 12 months (using subscales from the Global Appraisal of Individual Needs³⁰; 1=yes, 0=no); and 4) chronic health problems (1=yes, 0=no).

Dependent Variables

Completing the 3-month follow-up interview (regardless of follow-up strategy) is the dependent variable in the first logistic regression model (1=completed, 0=not completed). In the second model, follow-up strategy is the dependent variable (1=located using FB, 0=located using other means).

Data Analysis

Three sets of analyses were conducted using IBM SPSS Statistics for Windows, Version 21 (IBM Corp., Armonk, New York). First, chi-square and t-tests were used to explore differences between FB users and non-users. Second, a multivariate logistic regression model examined predictors of having completed a 3-month follow-up interview, with FB use as the variable of interest. Third, among FB users who had completed the follow-up interview, a multivariate logistic regression model examined predictors of being located using FB compared to other follow-up approaches.

Results

Sample Characteristics

The majority of the participants were white (98.9%) with an average age of 32.4. More than one-third (39.4%) reported being in a relationship and about half (49.7%) had completed high school. Only 23.9% were employed at least part time before incarceration.

Profile of Facebook Users

Almost two-thirds (64.4%; N = 284) of participants were FB users during the follow-up period. Bivariate analyses revealed that FB users were significantly younger than non-users (t(187.67) = 3.30, P = .001). FB users were also less likely to have chronic health problems (χ²(1, N = 284) = 5.64, P = .018), were more educated (t(182)= −2.18, P = .030), and were more likely to have used alcohol in the 30 days prior to incarceration (χ²(1, N = 284) =3.84, P = .050). FB users were also significantly more likely to be located for their 3-month follow-up interview (87.4% vs 63.4%; P < .0001). There were no differences in mental health history.

Follow-Up Completion

More than three-fourths (78.9%) of participants eligible for follow-up were successfully located using any method. A multivariate logistic regression analysis was conducted to examine FB use as a predictor of completing the follow-up. Analyses indicated that FB users had nearly 6 times the odds of being located for follow-up than nonusers (P < .0001), and experiencing symptoms of anxiety more than tripled the odds of being followed up (P = .005). Having a driver's license decreased the odds of completing follow-up by more than 50% (P = .035), and having been to prison during their lifetime decreased the odds of completing follow-up by more than 60% (P = .039).

Of those participants who completed the follow-up interview, more than half (57.1%) were located using FB. Among FB users who had completed the follow-up, more than three-fourths (78.1%) were located using FB. A second multivariate logistic regression analysis found that among FB users who had completed the follow-up, there were no significant predictors of being followed up using FB compared to being followed up using other approaches.

Discussion

This study contributes to the literature on the use of social media as a follow-up approach for out-of-treatment, hard-to-reach populations of rural drug users. Study findings indicated FB use was common among rural drug-using women, and that FB users were more likely to be younger and more educated, which is consistent with other literature.³¹ FB users were also more likely to be current alcohol users and although this likely did not impact follow-up via FB, it is possible that there is an underlying social aspect to both being on FB and alcohol use.

Study findings highlighted the utility of FB for locating traditionally hard-to-reach research participants in less densely populated areas. Past research suggests that economic disadvantage among rural populations often results in limited access to modern modes of communication^32,33 and reliable transportation,³⁴ posing challenges for study follow-up. These barriers become more pronounced when working with transient populations like drug users.³⁵ However, the current study found that among a sample of female offenders in rural communities, FB users were nearly 6 times more likely to complete the follow-up interview—suggesting that FB may offer a practical strategy for maintaining contact with geographically isolated and transient women for research follow-up over time. However, study findings also indicated that among FB users who were located and interviewed at follow-up, no other variables emerged as significant predictors of being followed up using FB compared to other follow-up approaches. In other words, although FB users were more likely to be located for a follow-up interview, there were no characteristics of FB users that predicted being followed up using FB versus other follow-up approaches.

There was also a high rate of mental health problems self-reported among both FB users and non-users, which is consistent with research showing high rates of mental health problems among rural, economically disadvantaged populations^36,37 and criminal offenders.^38,39 The regression specifically shows that women who self-reported symptoms of anxiety were at increased odds of completing follow-up, suggesting that women experiencing anxiety may also experience decreased mobility, making them easier to locate. Past research indicates that individuals with limited mobility are more likely to also experience anxiety.^40,41 Study results further support a connection between limited mobility and follow-up success since having a valid driver's license decreased the odds of follow-up.

Criminal history is also important for using FB for follow-up. While women were recruited from local rural jails, having served time previously in prison decreased the odds of completing a follow-up. Past research indicates that prisoners returning to the community often face barriers including stable housing,⁴² which potentially impacts study follow-up. Future research should continue to explore the relationship between criminal history and follow-up success, particularly how types of criminality impact follow-up.

Study limitations should be considered. First, this study explores FB use as a predictor of completing a 3-month follow-up interview. Additional research should include longer follow-up time intervals, particularly among hard-to-reach populations such as rural, drug-using women. Provided the challenges associated with long-term follow-up,³³ FB and other social networking sites offer a potential low-cost method for following up with research participants beyond 3 months.^43,44 Second, covariates were self-reported which is subject to recall bias and accurate self-disclosure. However, studies indicate that self-report data from criminal offenders and substance users are reliable and valid.^45-47 Third, FB use was also limited to whether participants were active FB users. Frequency of FB use and mode of access (eg, cell phone or personal computer) should be considered in future research. Fourth, data pertaining to Internet availability in participants’ communities were not collected. Although research has pointed to increased Internet adoption in rural communities, some rural communities still lack Internet access.⁴⁸ This should be considered in future studies. Finally, because this study was drawn from a larger study, it did not have an experimental design with a randomized comparison group. While this means the study is not a perfect test of the effectiveness of FB as a follow-up tool, study results do suggest that FB use is related to follow-up success and warrants further exploration.

Conclusions

Despite limitations, study findings suggest that Facebook is useful for locating rural research participants for follow-up. Specifically, this study shows that FB is widely used among rural drug-using women and suggests that FB may strengthen long-term follow-up among geographically isolated participants. Future studies should examine FB use among others, including the possibility of social media platforms for delivering interventions in less densely populated areas. In conclusion, it seems that FB will remain a robust option for following up and engaging with hard-to-reach study participants over time.

Table 1.

Bivariate Comparisons of FB Users and Non-Users (N=284)

	FB Users (n=183)	Non-Users (n=101)
Demographic Information
Age***	31.3	34.8
Race (% white)	98.4%	100.0%
Education (years completed)*	11.3	10.6
Currently in a relationship	36.1%	45.5%
Employed at least part time in 6 months prior to incarceration	21.3%	28.7%
Income in 6 months prior to incarceration	$9,011.15	$8,384.24
Currently have a valid driver's license	36.1%	31.7%
Substance Use History
Used alcohol in 30 days prior to incarceration*	32.8%	21.8%
Used any drugs in 30 days prior to incarceration	95.6%	93.1%
Injected any drugs in 30 days prior to incarceration	57.4%	49.5%
Criminal History
Age of first arrest	22.8	24.3
Number of times arrested	2.7	2.8
Number of times incarcerated as an adult	5.9	6.9
Ever been to prison	18.6%	12.9%
Mental & Physical Health
Major Depressive Disorder (past 12 months)	69.4%	69.3%
Generalized Anxiety Disorder (past 12 months)	46.4%	44.6%
PTSD (past 12 months)	68.9%	61.4%
Currently have a chronic health problem that interferes with your life*	25.1%	38.6%
Follow-Up
Followed up at 3 months (regardless of method)***	87.4%	63.4%

^*P ≤ .05

** P ≤ .01

^***P ≤ .001

Table 2.

Logistic Regression Models

	Model #1 (N=284)	Model #2 (N=160)
	Odds Ratio (95% CI)	Odds Ratio (95% CI)
FB User	5.87*** (2.85-12.08)	-
Age	1.02 (.96-1.08)	0.98 (.90-1.06)
Education	0.98 (.84-1.12)	0.92 (.75-1.12)
Currently in a relationship	1.77 (.87-3.63)	1.67 (.67-4.12)
Employed at least part time in 6 months prior to incarceration	1.21 (.54-2.74)	0.85 (.30-2.45)
Income in 6 months prior to incarceration	1.00 (1.00-1.00)	1.00 (1.00-1.00)
Currently have a valid driver's license	0.46* (.23-.95)	1.12 (.45-2.78)
Used alcohol in 30 days prior to incarceration	1.61 (.70-3.72)	1.47 (.60-3.63)
Injected any drugs in 30 days prior to incarceration	0.79 (.39-1.59)	1.28 (.54-3.02)
Age of first arrest	0.97 (.90-1.04)	1.06 (.95-1.18)
Number of times arrested	1.28 (.98-1.67)	0.85 (.60-1.22)
Number of times incarcerated as an adult	1.01 (.97-1.05)	1.22 (.99-1.51)
Ever been to prison	0.37* (.14-.95)	0.68 (.20-2.31)
Major Depressive Disorder (past 12 months)	0.74 (.31-1.77)	0.73 (.24-2.25)
Generalized Anxiety Disorder (past 12 months)	3.11** (1.41-6.83)	1.93 (.71-5.23)
PTSD (past 12 months)	0.95 (.42-2.16)	0.90 (.34-2.40)
Have a chronic health problem that interferes with your life	0.92 (.43-1.97)	0.84 (.30-2.34)

^*P ≤ .05

^**P ≤ .01

^***P ≤ .001

NOTE: Variables with limited variance (“white” and “used any drugs in 30 days prior to incarceration” were not included in logistic regression analyses.