The Impact of COVID-19 and Telehealth Services on Attrition Rates in Psychotherapy

Abstract

Attrition in psychotherapy has been identified as a significant obstacle in the productive delivery of mental health services. Defined generally as the ending of a treatment prior to proper optimal benefit, attrition both hinders treatment efficacy and costeffectiveness in therapy. With the demands for quality mental health services increasing, resources must be identified to reduce barriers to such services. The COVID-19 pandemic has resulted in the emergence of one potential resources: telehealth services. The current study aims to identify how COVID-19 and telehealth services have influenced attrition by analyzing attrition rates from both before and during the pandemic in a community health center where a transition to telehealth was made at the start of the pandemic. In addition, the variables of age, gender, socioeconomic status, and insurance coverage were also tested as potential predictors of attrition. Using de-identified patient information, clients who had participated in therapy services within a six-month period at a community health center (N = 329) were selected. A survival analysis was used to assess the time taken from initial appointment to the point of attrition. Results indicated that those who attended therapy via telehealth were less likely to stop attending treatment than those who participated in therapy in person. Individuals who used both in-person and telehealth visits were the least likely to terminate treatment prematurely. Clinical implications include the need for therapists to offer both telehealth and in-person services in order to give clients more resources to reduce a large barrier to needed mental healthcare treatment.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10591-023-09661-0.

Therapeutic outcomes are crucial for understanding the process of therapy and whether it creates change. In psychotherapy, clinicians are constantly focused on therapeutic outcomes and the change process as it allows them to assess their own effectiveness as therapists (Wampold, 2019). Currently, various models and methods of therapy exist with their distinct processes, views, and interventions, yet each aims to promote change in clients (Wampold, 2019). While research focused on change has shown general indications that clients do benefit from therapy (Erekson et al., 2018), studies comparing psychotherapy methods have not found any particular model to be more effective in the change process than another (Shadish & Baldwin, 2002, 2009; Wampold et al., 2016). With change being so important in therapy services, clinicians should look at all factors that may serve as barriers to the change process. One common barrier that has been identified among mental health service providers is attrition.

Attrition in psychotherapy is a significant obstacle in the productive delivery of mental health services as it hinders treatment efficacy and cost-effectiveness (Swift et al., 2012; Wierzbicki & Pekarik, 1993). The term attrition has been defined generally as the ending of a treatment prior to proper optimal benefit (Roseborough et al., 2016). Throughout the literature, synonymous terms to attrition have been used such as early termination (Bohart & Wade, 2013), premature discontinuation (Swift & Greenburg, 2012), and most popular, dropout (Barrett et al., 2008; Baruch et al., 1998; Fenger et al., 2010; Khazaie et al., 2016; Lopes et al., 2017). A meta-analysis looking at general studies with various definitions of attrition in therapy conducted by Swift and Greenberg (2012), found that the average rate of dropout was found to be 19.7%, implying that about one in five clients would terminate treatment prematurely. It is important that researchers and clinicians work to mitigate this issue as attrition greatly influences the individual client’s ability to change and the client’s access to mental health services in general. Anchored in Penchansky and Thomas’ (1981) expansion on Anderson’s Behavioral Model of Heath Services Use (1995), the purpose of this study is not only to identify possible factors that predict attrition but also to understand a specific resource (i.e., telehealth) that can increase access to meet the demand of mental health services.

Effects of Attrition

For the client, it can be assumed that prematurely ending therapy essentially interrupts therapeutic treatment which consequently diminishes the client’s rate of change (Xiao et al., 2017). While Lopes et al. (2017) states that attrition does not necessarily indicate clinical failure, he does note that in the long run, change will take much longer to occur in individuals who abandon treatment than those who complete it. Attrition and no-show rates also affect the service provider as they contribute to a loss of revenue, underutilization of time, and long waitlists (Barrett et al., 2008; Swift et al., 2012). Barrett et al. (2008) comments that the larger community may be impacted by nonattendance of therapy as it tends to drain limited mental health resources for the public. With there already being a lack of mental health professionals, attrition rates only increase the need for service providers. It is because of these vast effects that much of the research on this topic focuses on the factors that may predict attrition.

Although the data are not consistent in the literature on predictors of attrition (Bohart & Wade, 2013), factors including level of education, socioeconomic status, and life/environment circumstances have shown significant results across multiple studies. Clients reporting higher education (college degree) were more likely to remain in treatment when compared to those who reported only a high school or vocational schooling as their highest level of education (Barrett et al., 2008; Fenger et al., 2010; Roseborough et al., 2016). Attrition was also common among those who had a lack of insurance coverage, financial problems, or were of a lower socioeconomic status (Barrett et al., 2008; Khazaie et al., 2016). In multiple studies, results also indicated that many clients unexpectedly miss appointments or drop out entirely due to circumstances of life such as physical illness, work conflicts, lack of transportation, and difficulty locating childcare (Barrett et al., 2008; Defife et al., 2012).

Attrition rates may also be dependent on the variation of clients within the therapy room. Hamilton et al. (2011) noted that both the profession of the provider and the modality of therapy (e.g., individual, couple, family) are predictors of attrition. However, the literature in this regard is not very clear. Some studies have found highest rates of attrition to be found among family therapy clients (Hamilton et al., 2011), others among couples (Werner-Wilson & Winter, 2010), and even some have found no significant difference across modalities (Masi et al., 2003). This inconsistency indicates that further research should be done to look at the effects of attrition on family and couples therapy clients. While the setting for this study did provide therapy to couple and family therapy clients, the data used did not distinguish between the modalities and could not consider therapy modality as a predictive factor.

With that being said, in general, there is a lack of consistency within the literature on the topic of attrition in psychotherapy. Much of the research that has been done is filled with confounding findings, replication failures, and relatively small differences between those who continue therapy and those who terminate prematurely (Wierzbicki & Pekarik, 1993). It is only through more examinations of attrition that such issues can be resolved. With many behavioral health organizations seeing as much as a 52% increase in the public need for services (Majlessi, 2020), it is imperative that more be done to understand the attendance of clients and reduce attrition rates so appropriate resources can effectively meet such demands.

Coronavirus Pandemic and Mental Health Accessibility

Access to treatment and quality care was made increasingly difficult when the coronavirus (SARS CoV 2) was declared a pandemic in March 2020 and created havoc as it spread across the globe (Malathesh et al., 2020). Since that time, nations around the world have reported elevated rates of anxiety, depression, stress, suicide risk, and post-traumatic stress as fears of contamination and quarantining have become a part of everyday life (Cooke et al., 2020; O’Connor et al., 2020; Wang et al., 2020). The pandemic has both increased the demand of mental health services as well as disrupted and halted many mental health organizations and the delivery of face-to-face mental health services in general (World Health Organization, 2020). During such a period where therapy services are critical, attrition poses an even greater danger to therapists trying to meet the growing mental health needs of their communities.

While it is difficult to identify specific causes behind attrition, therapists can utilize resources that may allow them to be more accommodating to clients and potentially increase rates of attendance. Telehealth is one such resource that is on the rise with the widespread availability and popularity of technology (Vockley, 2015). With the recent global pandemic, many clinicians have been forced to transition therapy sessions to be done via telehealth allowing them to continue to meet their client’s needs while ensuring medical safety and keeping physical-distancing requirements (Taylor et al., 2020). Such a dramatic shift in modality raises the question of how the pandemic and the availability of telehealth as a resource has influenced psychotherapy attrition rates.

Teletherapy as an Attrition Reduction Strategy

The COVID-19 pandemic has impacted in-person services significantly as stay-at-home orders and social distancing guidelines have been put in place to reduce spread (Taylor et al., 2020). As a result of the pandemic and the steady increase in technology, teletherapy is becoming increasingly popular and useful (Pickens et al., 2020). Skeptical clinicians initially thought teletherapy to be ineffective and unethical, however, evidence currently suggests that providing therapy through this modality has the same, and occasionally higher, levels of efficacy as face-to-face therapy (Twist & Hertlein, 2017).

In a systematic review, Turgoose et al. (2017) found that teletherapy methods were just as effective in reducing symptoms of PTSD in veterans than in-person methods. Both telephone-delivered and videoconferencing technology have been identified as supported treatments for psychologists treating clients with depression, anxiety, PTSD, or adjustment disorders (Varker et al., 2019). Burgoyne and Cohn (2020) found that telehealth can also serve as viable resource when seeing relational clients as it allows for more members of the system to participate in treatment. When surveying their clients, Burgoyne and Cohn found that 86% of clients and 80% of staff found teletherapy to provide good quality of care.

Teletherapy serves as a resource as it helps to improve access to mental health treatment by increasing availability, accommodation, affordability, and acceptability (Penchansky & Thomas, 1981). While issues such as service errors or other technical issues can affect a therapist’s ability to join or establish a relationship with the client electronically (Twist & Hertlein, 2017), teletherapy gives clients the opportunity to overcome many of the obstacles mentioned above (Wrape & Mcginn, 2018). The dimension of availability is increased as it enables clients to do it from home. Accommodation is enhanced as clinicians are more flexible with doing therapy in person or via telehealth depending on the client’s needs. Telehealth services can capture the domain of acceptability as they give clients more options of therapists across the country that may be more comfortable with their immutable characteristics (i.e., ethnicity, sex, social status). Affordability is even increased as it does not cost the client any extra to do so and potentially saves money in travel fees. It should be noted that such a resource is dependent upon having the necessary equipment. Although widespread technological advances make telehealth appointments possible for many communities, telehealth programs require adequate broadband access which may not be available for many rural and underserved populations (Hirko et al., 2020). With teletherapy being a relatively new resource, more research is needed in order to gain a broader comprehension of its benefits and explore how such services can reach more underprivileged communities.

Purpose of the Study

Attrition in psychotherapy is problematic as it negatively impacts therapy clients, clinicians, and those awaiting mental health services in communities (Barrett et al., 2008). Attrition virtually stands as a barrier to the rising demands of mental health services. In order to effectively reduce such effects, appropriate resources need to be implemented in mental health organizations to increase availability, accommodation, affordability, and acceptability for clients. The purpose of the present study was to investigate the influence of the COVID-19 pandemic and teletherapy use on the rates of attrition in mental health services.

Following Penchansky and Thomas’ model of healthcare access, it is hypothesized that teletherapy will increase access to mental health treatment and that rates of attrition will have decreased since COVID-19 and the switch to teletherapy. This will indicate that the risk of drop out is lower for teletherapy than in-person. In accordance with previous literature (Barrett et al., 2008; Khazaie et al., 2016), it is also hypothesized that variables such as insurance coverage and SES will be significant indicators of dropout risk while variables of age and gender will not predict significant risk. Confirming these hypotheses will provide needed information to the literature on attrition as it will indicate that telehealth can serve as a resource. Knowledge of such a resource could benefit access to mental healthcare as clinicians utilize it to reduce attrition rates. Examining specific predictor variables could indicate which demographics in particular might need such resources as they may be more likely to drop out of treatment prematurely.

Methods

In this study, existing data from previously scheduled psychotherapy appointments at a community health center were investigated. The following sections will elaborate more on the setting where the data were collected, the participants included, and the procedure used in acquiring the data prior to analysis. In addition, the analytic strategy will be detailed along with figures to explicitly illustrate what was done to yield the acquired results.

Setting

Data in this study were collected from a Federally Qualified Health Center (FQHC) with seven locations across the Western United States. This community health center provides medical, behavioral health, dental, and pharmaceutical services to the community, particularly to those of lower socioeconomic status as they offer a sliding fee scale for payment. The sliding fee scale is based off of household size and income and is divided into four levels: Level 1 (up to 100% of federal poverty level), Level 2 (up to 133% of federal poverty level), Level 3 (up to 150% of federal poverty level), and Level 4 (up to 200% of federal poverty level). Within the United States, the federal poverty level is a standard measure containing specific income thresholds below which individuals and families are considered to be poor (Betson & Michael, 2000). Such thresholds are adjusted each year to account for inflation. This sliding fee scale system was used as a method of measuring income and socioeconomic status (SES) in this study as it utilizes client’s tax returns, pay stubs, and bank statements to accurately compute discount qualification.

Mental health providers across all of the clinics consisted of three Licensed Clinical Social Workers (LCSW) and one Licensed Marriage and Family Therapist (LMFT). Providers offer therapy services to individuals, couples, and families in 45-min sessions. Therapy services were offered via face-to-face and telehealth until April 1, 2020 when the organization transitioned to telehealth services as their primary modality due to the COVID-19 pandemic.

Participants

The sample of this study consisted of 329 clients receiving mental health services at a FQHC in the Western United States. By nature, clients attending FQHC’s often experience higher levels of stress due to their financial situation, cultural barriers, or other life circumstances. Many of the individuals included in the sample of this study suffered from mental health issues of higher severity due to their inability to receive services elsewhere. Participants were selected if they had been seen for therapy between the dates of January 1 and June 30, 2020, in order to examine the attrition rates both before and during the COVID-19 pandemic. In cases where a family or a couple was being treated for therapy, not every individual was tracked for attrition, only the identified patient (the individual whose name is on the schedule). Clients were excluded if they were already participating in teletherapy prior to the health center’s transition to exclusive telehealth services.

Demographics of the sample varied with 135 (41%) being male, 193 (58.7%) being female and 1 (0.3%) transgender male. Race of participants consisted of 302 (91.8%) White, 14 (4.3%) Hispanic, 3 (0.9%) Black, 2 (0.6%) American Indian, 1 (0.3%) Asian, and 7 (2.1%) not reported. Average age of the participants was 32 with 266 (81%) being adults and 69 (19%) being minors (below the age of 18). Between January 1 and March 31, 2020, 75 (22.8%) clients attended therapy and were considered in the “in-person” group. Between April 1 and June 30, 2020, 64 (19.4%) clients attended therapy solely via telehealth, placing them in the “telehealth” group. From January 1 to June 30, 2020, there were 190 (57.8% clients that attended therapy through both “in-person” and “telehealth” platforms. Full descriptive statistics are presented in Table 1.

Table 1.

Demographic summary of sample by modality of therapy

	In-Person N = 75	Both N = 190	Telehealth N = 64	p*
	M (SD)	M (SD)	M (SD)
Age, years	30.17 (17.14)	32.58 (15.74)	36.27 (14.69)	0.078
Therapy duration, Weeks
Prior to baseline	21.31 (40.85)	65.18(71.30)	0 (0)	–
Baseline to endpoint	4.95 (4.08)	11.25 (3.39)	4.24 (4.10)	< 0.001
	n (%)	n (%)	n (%)
Gender				0.026
Male	41 (55%)	70 (37%)	24 (38%)
Female	34 (45%)	119 (63%)	40 (62%)
Unknown		1
Insurance				0.035
Government	21 (28%)	50 (26%)	16 (25%)
Private insurance	40 (53%)	125 (66%)	35 (55%)
Slide	12 (16%)	12 (6%)	13 (20%)
Out of pocket	2 (3%)	3 (0.2%)	0 (0%)
Income				0.231
Level 1	21 (30%)	60 (32%)	16 (25%)
Level 2	7 (8%)	10 (5%)	8 (12%)
Level 3	1 (4%)	8 (4%)	4 (6%)
Level 4	4 (10%)	22 (12%)	6 (9%)
Unknown	42 (49%)	90 (47%)	30 (47%)
End Point for 6-month Window				< 0.001
Dropped out of therapy early	67 (89%)	38 (20%)	20 (31.2%)
Completed therapy course	8 (11%)	8 (4.2%)	3 (5%)
Continuing therapy (censored)	–	144 (76%)	41 (64%)

Income level 1 is the lowest and 4 is the highest based on Table 1 in the Online Appendix

*Significance (p) is indicative of an analysis of variance (ANOVA) for continuous variables and Chi-squared test of independence for each categorical variables with therapy modality

Procedure

Clients participated in therapy as usual and typically attended therapy once a week. On April 1, 2020, the community health center changed their therapeutic delivery exclusively to telehealth services in response to the COVID-19 pandemic. Clients were automatically considered as having terminated prematurely if they only attended an intake and had no future appointments scheduled. Attrition was defined by the therapists and patient charts previously signed by the therapy provider were utilized to verify premature termination. Attrition classified by the therapist’s clinical judgement has historically been found to be a preferable operationalization as the concept of dropout in and of itself stems from the clinician’s judgment that clients terminate inappropriately from therapy (Hatchett & Park, 2003; Pekarik, 1985; Swift & Greenberg, 2012).

Participants whose clinical notes stated that a follow up was recommended but stopped attending were considered dropouts. De-identified patient information was used so no consent was required by clients.

Variables

The main variable of interest in this study was the delivery of therapy: in-person, telehealth, or both. Other potential confounding factors examined included age, gender, socioeconomic status (SES) via sliding fee scale brackets, and insurance coverage (private insurance, federal health programs, sliding fee scale, out-of-pocket). While race and ethnicity were originally thought to be a confounding factors, a lack of diversity in the sample inhibited such factors from being taken into consideration.

Analytic Strategy

Survival analysis, also called time-to-event analysis, was used as the primary strategy of analysis. Survival analysis focuses on the expected duration of time until occurrence of an event of interest (Kleinbaum & Klein, 2010). In this case, such an event was attrition in psychotherapy. Such a strategy can assess the time taken from the client’s initial appointment to point of premature termination. Survival analysis allows for the analysis of staggered entries by moding each individual’s time since entry, meaning that although each client’s initial appointments were all at different calendar dates, they can still be interpreted in regard to the same context. Since attrition is not experienced by every client, those for whom the event did not occur during the observation period were considered “censored” having an unknown outcome. Right censoring takes place when the time of the event is known to be greater than some value but the time it took for attrition to occur after the observation period is unspecified/unknown. This was important as it allowed for the inclusion of data for clients who were still continuing in therapy, those who graduated, and those who were referred out after the designated observation period.

Thus, within the 6-month window of observation, there are three possible end points: early termination (drop out), successful completion of therapy, or continuation of therapy (censoring). The observed time, in weeks, for each participant is measure from their individual baseline date till their terminal endpoint date. Figure 1 displays ten participants in each group to show examples of baseline and endpoint combinations observed in the 6-month window of the study. Baseline is defined as the first therapy appointment date or January 1, 2020, whichever came later, for the In-Person Modality participants. For participants who started or switched to Telehealth, baseline is defined as the first therapy appointment date or April 1, 2020, whichever was later. Duration of therapy preceding baseline was captured and entered as a covariate during analysis.

Fig. 1.

Visual presentation of example observed time therapy duration and end points

Implementation of Kaplan–Meier (KM) plots (seen in Fig. 2) provided a descriptive graphical presentation to compare the different populations (all in-person, in person and telehealth, all virtual) and illustrate the attrition rate of each group. This non-parametric curve visually represents the survival rate across time where the survival probability drops vertically whenever one or more events of interest (attrition) occurs within the time interval (Kleinbaum & Klein, 2010).

Fig. 2.

Kaplan–Meier plot for therapy modality

In conjunction with the KM plot, a log rank test compared the population groups to see if a statistically significant difference exists between them. Specifically, this tests the null hypothesis that the survival curves from the KM plot are identical over time. The log rank test, however, does not allow for comparisons across continuous variables such as age unless they are first discretized (age in years converted to categorize such as under 18, 19–30, and over 30), resulting in loss of information. Another drawback is that KM plots and log-rank test lack estimates of effect size (Kleinbaum & Klein, 2010).

The Cox proportional hazards regression (Cox Regression) allows for quantification of differential risk across both categorical and continuous variables. Such a model provides a hazard ratio, which explains the probability of the event (i.e., attrition) occurring at any point in time for therapy type (in-person, telehealth, or both), while simultaneously controlling for any number of covariates (potential confounding factors). This directly answers the research question of how each type of therapy will influence the risk of attrition. All analyses were conducted in R 4.2 (R Core Team, 2022) and utilized the survival (Therneau, 2022) package.

Results

In total, one-hundred and twenty-five clients from the total sample (N = 329) dropped out of therapy during the six-month period, equaling 37.9% of the total sample. Within each individual group approximately 89.3% of the in-person group, 20% of those in the both group, and 31.2% of those in the telehealth group had terminated prematurely at the end of six months. Means, medians, and standard deviations for the groups can be found in Table 2. Overall, the mean survival time for the entire sample was about 9.57 weeks. The average survival rates for each individual modality group were 5.40 weeks for in-person, 11.45 weeks for both, and 8.67 weeks for telehealth.

Table 2.

Means and medians for survival time, weeks by modality

Therapy modality	Mean
	Estimate	SE	95% CI		Estimate	95% CI
			Lower bound	Upper bound		Lower bound	Upper bound
In Person	5.40	0.48	3.86	7.00	5.000	3.86	7.0
Both	11.45	0.24	NA	NA	–	–	–
Telehealth	8.67	0.75	6.71	NA	–	–	–
Overall	9.57	0.26	NA	NA	–	–	–

Hypothesis 1

A KM plot was used to illustrate the survival function of each therapy modality. The curve illustrates that the ‘both’ group maintained a higher survival (non-attrition) rate in comparison to the telehealth and in-person groups. At 12.5 weeks the rate of survival was 59% for the telehealth group, 4% for the in-person group and approximately 80% for the ‘both’. A log rank test determined there were significant differences in the survival distribution for the different therapy modalities: in-person, telehealth, or both. The difference in the survival distributions were found to be statistically significant, χ²(2) = 168, p < 0.001 (Fig. 2).

Hypothesis 2

The Cox Regression model (See Table 3) found none of the predictors (age, gender, 4-level insurance coverage, 4-level income) to be statistically significant with therapy modality. However, even when controlling for such covariates, the model showed the therapy modality groups to be statistically significant from one another. The hazards ratio (HR) or exponentiated parameter estimate, compares both the in-person and the both groups to the telehealth group as statistical software defaults to using the last variable as the indicator variable. Payment not supplied by government assistance was associated with 60% higher rate of drop out at any given time. After controlling for government payment and the protective effect of a longer pre-baseline duration of therapy, the risk of drop out did vary by group. The in-person only participants were more than four times as likely to drop out of therapy than those in the telehealth only group, HR 4.61, p < 0.001. Participants who attended both modalities of therapy, were at a 60% lower risk of dropping out after the switch to telehealth as compared to participants that started with telehealth, HR 1.60, p = 0.031.

Table 3.

Cox regression of time to dropout (weeks)

Variable	Est	SE	z	df	p	HR
Therapy modality (ref = Telehealth Only)				2
In-person only	1.53	0.28	5.40	1	< 0.001	4.60
Both: Telehealth after in-person	− 0.89	0.31	− 2.89	1	0.004	0.41
Payment
Government vs. Not	0.47	0.22	2.15	1	0.031	1.60
Prior weeks (ref = none)
Less than 3 months	− 0.67	0.29	− 2–0.37	1	0.018	0.51
More than 3 months	− 1.01	0.22	2.15	1	< 0.001	0.37

Est. estimated beta parameter, HR hazard ratio

Discussion

The main objective of the current investigation was to add to the current literature of attrition rates in psychotherapy. More specifically, this study aimed to explore telehealth as a resource for accessibility and analyze its influence on attrition rates during the COVID-19 pandemic.

Treatment Modalities

Analyses revealed that the three modalities of therapy resulted in statistically different rates of attrition. The total attrition rate of the entire sample was approximately 38% which varies greatly from Swift and Greenberg’s (2012) metanalysis which estimated a rate of 19.7%. In other words, these results indicate that about two out of every five clients in our sample ended treatment prematurely. While this is a comparatively higher rate, the fact that the data were collected from a lower income population at an FQHC does reflect the previous literatures claims of attrition being higher among those of lower socioeconomic status (Barrett et al., 2008; Khazaie et al., 2016). As noted previously, the way attrition was defined in this study (i.e., therapist’s discretion) could have greatly impacted the resulting attrition rate and is a definite limitation which will be discussed later in more depth. The COVID-19 pandemic could have also played a major part in this increased attrition rate. Sickness from and fear of the pandemic could have served as a major barrier that inhibited therapy retention. Overall, the context surrounding the sample and the sample itself may be too different for the total attrition rate found to be generalized to other populations.

Comparing the different groups, the participants that had received psychotherapy both in person and via telehealth showed much lower odds of attrition when compared to the groups of just telehealth. Through Penchansky and Thomas’ model (1981), we can imply that having both telehealth and in-person therapy as options allows clients to find the “best fit” for each of the domains of availability, affordability, accommodation, and acceptability. By having both options available, clients may also be more likely to attend their appointments as they can adapt their delivery of therapy based on their current situations or potential barriers.

When looking strictly between the in-person and telehealth groups, results confirmed that those who were just in-person were over four times as likely to drop out than those who purely saw their therapists via telehealth. Such results supported the hypothesis that when teletherapy was offered as an option, attrition rates were reduced. In addition, while this study did not explicitly examine the treatment efficacy of telehealth services, these results strengthen the literature that shows teletherapy as an effective treatment option (Burgoyne & Cohn, 2020; Turgoose et al., 2017; Twist & Hertlein, 2017).

It can be inferred that participants attending in-person visits are more likely to stop attending treatment because such visits may present additional barriers when compared to the telehealth or both in-person and telehealth options. As noted in the literature review, unexpected life circumstances (e.g., illness, work conflicts, childcare) are often barriers to accessing mental health treatment (Barrett et al., 2008; Defife et al., 2012). When such circumstances arise, attending services at a mental health service provider is often not a possibility, whereas telehealth services provide a way for clients to still receive services from various locations, even with such confounding factors.

Demographic Variables as Predictors

The analyses run did not indicate any significant association between attrition and age, gender, SES, or insurance type as predictors for attrition, other than government payment. While the variables of age and gender were expected to result as insignificant predictors in accordance with previous literature (Bohart & Wade, 2013), the predictors of SES status and insurance coverage differed from the hypothesized results. As with earlier studies (Barrett et al., 2008; Khazaie et al., 2016) when controlling for the other variables, lower SES status and lack of insurance coverage were expected to be significant predictors of attrition. The Cox Regression model, however, did not support this expectation.

Although the results did not reflect what was hypothesized, the insignificance of such demographic variables does not differ greatly from the literature which has found inconsistencies in these variables as predictors (Bohart & Wade, 2013). However, the fact that the sample was collected from an overall lower income population may have influenced the outcome for this hypothesis. Due to this sample’s lack of variability, it cannot definitively be assumed that these demographic variables had no effect on attrition rates.

Clinical Implications

Knowing that attrition can greatly affect the rate of change for clients in therapy (Wierzbicki & Pekarik, 1993; Xiao et al., 2017), clinicians should take preventative measures to help limit barriers to mental healthcare access. With the results in this study showing higher survival rates in the ‘both’ and ‘telehealth’ groups, mental health providers should be aware of the potential resource that telehealth services can provide to reduce attrition. While some therapists may have concerns about teletherapy negatively impacting the therapeutic relationship (Twist & Hertlein, 2017), clinicians should take into account the benefits implied from the results in this study. By providing both the options for telehealth and in-person therapy delivery to clients, therapists may decrease their daily no-show rates and the number of last-minute cancellations. Providing telehealth services may be particularly beneficial for Marriage and Family Therapists as it can provide another avenue to involve more members of the system in treatment (e.g., a partner out of town, sick family member, parents living in another part of the state; Burgoyne & Cohn, 2020).

Therapists may also increase attendance for clients with specific diagnoses by having telehealth as an option. Those diagnosed with depressive disorders, panic disorder, social anxiety disorders, or other anxiety disorders could particularly feel more comfortable to begin services via telehealth as they can access such services while avoiding distressing circumstances such as social interactions or the outside environment (Wiederhold, 2020). Once adequate rapport has been established, the therapist can transition to in-person treatment or utilize other interventions via teletherapy to help clients confront such cognitions (Abramowitz, 2019). Lastly, in accordance with previous research, these results provide greater reasoning for therapists in rural areas to use telehealth and reach those that they might not have been able to if working strictly in-person (Wiederhold, 2020).

Limitations

A number of limitations are present within this study and should be considered when interpreting the findings. First, the use of extant data often constrained the study to the information that was initially collected as no further data could be collected prior to analysis. While the timeframe of six months provided significant outcomes, a longer time period (additional months prior to and after COVID) could have provided a bigger picture of the phenomenon studied. This extant data also limited which additional variables were available for consideration for this study. Items that have previously been found to be possible predictors of attrition such as the training of the clinician, DSM-V diagnosis, and modality (e.g., individual, couples, family) could not be controlled for or assessed because of the available data. Including patient’s diagnoses and health conditions (chronic vs acute) was initially considered, but the diagnoses of such items in the patient’s charts were too inconsistent to be beneficial and thus were not taken into consideration.

The FQHC utilized in this study was expected to have sufficient diversity to be representative of various racial and ethnic backgrounds. Race and ethnicity were examined as other predictor variables, but as the sample was predominately white (91.8%), neither could be used as significant predictors. While it appears that the other services provided at the FQHC (e.g., medical, dental, pharmacy), served clients of non-white backgrounds, the data collected of behavioral health services seemed to reflect more of a white majority. Racial and ethnic variations of attrition rates have previously been found and should be considered to better represent and understand the experiences of minority populations (Green et al., 2020). This absence of diversity inhibits the ability to generalize these findings to other populations.

An unexpected limitation was the predominance of clients who were seen both before and after the start of the pandemic (n = 190) in comparison to those who only received therapy via telehealth (n = 64) or in-person (n = 75). The makeup of the ‘both’ group largely consisted of clients who had been seeing a therapist for a long period of time prior to the dates assessed. Naturally, those clients who have an established history with their therapist are more likely to continue treatment when compared to those who just start treatment. This confounding factor could have contributed to the low attrition rate of the both group when compared to those who had only participated in therapy via telehealth. While it was beyond the scope of this study, longevity of treatment is a potential confounding factor and should be considered in future studies regarding attrition.

The retrospective data used in the study limited the ability to control for various confounding variables as well. With this data, clients could not be inquired as to what influenced them to drop out of treatment (e.g., life stressors, dissatisfaction with treatment or the therapist, financial issues, believing they no longer needed treatment). It is because of this inability to interact with participants that it cannot be concluded that telehealth services alone influenced attrition in each group.

Lastly, the manner in which attrition was defined in this study is also, in and of itself, a limitation. Therapist’s discretion in their clinical notes about whether or not a patient should continue treatment was chosen because of its easy accessibility with the data. However, as noted by Wierzbicki and Pekarik (1993), this judgement can often lack reliability as each therapist may use different criteria to define who is or is not ready to terminate treatment.

Future Directions

A prospective study that was not limited by the given data would allow for an examination of the impact of telehealth services on attrition during a longer period of time, rather than just the six-month period that we were given. Future studies using this or similar data should analyze longevity of therapy as a potential predictor variable. Additionally, similar studies should be conducted using more diverse samples (i.e., representative of sexual, racial, and ethnic minorities) from other FQHC’s across the United States in order to allow the results to become more generalizable. By doing so, race and ethnicity could be considered as a variable and potential results could indicate which racial/ethnic groups may have more barriers to accessing mental healthcare treatment.

Future research in this field may also consider studying the effect of an established therapeutic relationship on attrition rates. As mentioned above, many of those that did not dropout in this study appeared to have started therapy much earlier than when the data were collected, allowing them to have a longer history with their therapist. The therapeutic alliance, or the relationship the therapist has with the client, has been found to be one of the most important therapist-influenced conditions for client outcomes (Fife et al., 2013). While this study did not have the means to analyze the therapeutic relationship as a factor for attrition, prospective studies in attrition may find it to be especially influential.

To strengthen the statement that utilizing both in-person and telehealth services can serve as protective factors against attrition in therapy, similar studies should be conducted after the COVID-19 pandemic. Doing so would eliminate the pandemic as a confounding factor and potentially provide a clearer understanding of the relationship between therapy modality and rates of attrition. With services via telehealth being a choice rather than a mandate post-pandemic, interaction with the therapy clients through surveys would also be preferred to see what factors influence their choice in determining how they would like therapy to be delivered. Identifying such factors could further strengthen the use of the Healthcare Utilization model and its dimensions of access.

Conclusion

Previous research has shown attrition to interfere with access to productive mental healthcare delivery as well as diminish a client’s rate of change (Wierzbicki & Pekarik, 1993; Xiao et al., 2017). In exploring telehealth as a resource to reduce attrition rates, it was found that having both telehealth and in-person therapy visits as an option for treatment may decrease the likelihood of clients terminating treatment prior to optimal benefit. While no significant predictor variables were identified to explain why attrition occurred, these results give insight as to how clinicians can provide clients with a resource to prevent attrition. Such findings not only confirm previous conclusions made about attrition but also enhance the current literature on the subject. Adding these results to the available literature on attrition provides a resourceful solution to reducing rates rather than focusing primarily on the root cause of premature termination. While this study has significant limitations and does not signify specific outcomes, the implications of these findings suggest that future research should examine the benefits of telehealth resources and their possibility of providing greater mental healthcare access.

Supplementary Information

Below is the link to the electronic supplementary material.

Data Availability

The datasets analyzed during the current study are owned by Bear Lake Community Health Centers, but restrictions apply to the availability of this data, which were used under contract from Bear Lake Community Health Centers, and so are not publicly available. Data may be available from the authors upon reasonable request and permission from the Bear Lake Community Health Centers.

Declarations

Conflict of interest

No potential financial or non-financial conflicts of interest were identified in this study.

Ethical Approval

The research was submitted to IRB but was exempt because it was extant data with no identifiers.

Informed Consent

Due to the fact that all data were de-identified, informed consent was not required.