Search
Member Login

Which Clients Benefit Most from In-Person Versus Telehealth Therapy?

by: Wilson Trusty and Rebecca Janis
Category: New Findings
Nov
25

Telehealth therapy has become a permanent addition to many college counseling centers since the beginning of the COVID-19 pandemic. Research on telehealth has found that clients, on average, make equivalent improvement in symptoms and form similar working alliances in telehealth (TH) and in-person (IP) care in counseling centers and other settings ​(CCMH Blog 2023; Davis et al., 2023; Gurm et al., 2023; Lalor et al., 2023)​. However, less is known about whether certain types of clients benefit from one modality more than the other. This is important because results from studies of the “average client” may not always apply to a specific client a therapist is treating. Identifying whether certain clients benefit more from TH or IP psychotherapy could inform counseling center clinicians’ and administrators’ decisions about which students might benefit more from either modality. As such, this blog examined three questions about TH and IP therapy in college counseling centers: 

  1. Do TH and IP therapy produce different amounts of symptom change for different clients? 
  2. What types of clients benefit more from TH and IP therapy? 
  3. Can we improve clients’ outcomes by assigning them to the modality (TH or IP) predicted to be best for them? 

To answer these questions, we used data from 17,416 clients receiving individual therapy at 138 college counseling centers in the United States from 2021 to 2024. We defined symptom change as the difference between clients’ first and last CCAPS-34 administrations during therapy, and client characteristics were gathered from the Standardized Dataset Client Information Form that clients completed at the beginning of treatment.  For a complete list of the characteristics evaluated, see the Appendix. 

Data analysis was conducted in three steps. First, a formula was derived from client outcomes in the national sample to predict the amount of CCAPS change clients were likely to achieve in both TH and IP therapy based on their initial symptoms, treatment history, suicidal and homicidal risk, and demographic/identity-related characteristics. Next, we determined the extent to which each client was predicted to achieve more change in TH or IP and whether client characteristics could inform the modality in which they would experience more symptom improvement. Finally, we calculated how much additional CCAPS change each client would have achieved if they were assigned to their optimal treatment modality. Details on these analyses are in the appendix at the end of the blog.  

Do TH and IP therapy produce different amounts of symptom change for certain clients? 

We found that the majority of clients were predicted to experience similar amounts of symptom change on the CCAPS regardless of whether they received TH or IP therapy. This is illustrated in the figure below. In the figure, negative numbers favor TH (i.e., indicating that a client would experience more improvement in TH), while positive numbers favor IP. The fact that the majority of clients are clustered near zero shows that the two modalities have nearly equivalent outcomes for most individuals. However, there was a small proportion of clients who were predicted to experience slightly more symptom improvement in one modality versus another. 

​

What types of clients benefit most from TH versus IP therapy? 

We next tested whether clients’ symptoms when beginning therapy, suicidal and homicidal risk, treatment history, and demographic/identity-related characteristics predicted whether TH or IP would be the optimal service mode for them. The full list of variables tested is included in the Appendix . Overall, we found that these client characteristics only accounted for approximately 1% of the variability in symptom change between TH and IP. That is, none of the client characteristics we examined could meaningfully predict whether they would benefit more from IP versus TH therapy. This suggests that clients were served equally well in TH and IP services regardless of their pretreatment characteristics.   

Can we improve clients’ outcomes by assigning them to the modality (TH or IP) predicted to work best for them? 

Finally, we examined how much clients would improve if they received their "optimal" type of therapy (telehealth versus in-person), as determined by the prediction formula created in the previous steps. The results revealed both TH and IP services led to improved symptoms, but assigning clients to their optimal treatment mode did not meaningfully improve symptom change on the CCAPS Distress Index (DI), as the 0.03 difference in change in the graph below is not clinically significant. This suggests that counseling centers are able to help most clients benefit equally through telehealth or in-person therapy regardless of their symptoms, risk to themselves or others, demographic characteristics, and treatment history. 

 

​Note. DI: CCAPS Distress Index. 

Takeaways 

Summary 

  • Consistent with past research, our results show that the individual therapy services provided by college counseling center clinicians are equally effective in TH and IP modalaties. 

  • The small differences in how much clients were predicted to benefit from TH or IP therapy were not meaningfully related to most individuals’ initial symptoms, risk to themselves or others, demographic characteristics, or treatment history. 

  • Assigning clients to their “optimal” treatment modality based on their characteristics was predicted to have negligible additional benefits in terms of symptom reduction. 

  • Taken together, these results suggest that counseling centers are equally effective with a wide range of students regardless of whether they provide therapy via TH or IP modalities. 

Clinical Implications 

  • Since the present results show that counseling centers provide effective services via TH and IP modes of care, we recommend that centers and institutions make both modalities available to their students when possible. This is especially important since past research highlighted that TH and IP therapy facilitate access to care for different types of  clients (e.g., sexually diverse students are more likely to be seen via TH, but students living with roommates are more likely to use IP therapy; Trusty et al., 2025). Thus, preserving access to both modalities could help centers treat a wide range of students without sacrificing the effectiveness of therapy. 

  • While this study showed that clients’ pre-treatment characteristics were largely unrelated to whether TH or IP therapy would work best, there are other characteristics we did not measure, such as clients’ treatment preferences and presenting concerns. Additionally, we did not test whether particular combinations of characteristics predicted better response to TH or IP treatment. Because of this, it is important for counseling centers to integrate these findings with their clinical judgment regarding individual clients’  circumstances and their centers’ capabilities when deciding which modality is appropriate for each client. 

  • It is important to note the current investigation only compared TH and IP therapy provided by counseling centers. We did not include services delivered by telehealth vendors to college/university students. As such, these results do not evaluate the effectiveness of services provided by telehealth companies. If telehealth vendors’ services are being considered by a college/university, we encourage a robust collaboration with the local college counseling center to evalute the need and establish the system of access and care that will be implemented by the company. 

References 

Davis, K. A., Zhao, F., Janis, R. A., Castonguay, L. G., Hayes, J. A., & Scofield, B. E. (2023). Therapeutic alliance and clinical outcomes in teletherapy and in-person psychotherapy: A noninferiority study during the COVID-19 pandemic. Psychotherapy Research, 34(5), 589–600. https://doi.org/10.1080/10503307.2023.2229505 

Gurm, K., Wampold, B. E., Piatt, C., Jagodzinski, R., Caperton, D. D., & Babins-Wagner, R. (2023). Effectiveness of telemental health during the COVID-19 pandemic: A propensity score noninferiority analysis of outcomes. Psychotherapy, 60(2), 231–236. https://doi.org/10.1037/pst0000472 

Lalor, I., Costello, C., O’Sullivan, M., Rice, C., & Collins, P. (2023). Brief psychological interventions in face-to-face and telehealth formats: A comparison of outcomes in a naturalistic setting. Mental Health Review Journal, 28(1), 82–92. https://doi.org/10.1108/MHRJ-05-2022-0029 

Trusty, W. T., Scofield, B. E., Cooper, S. E., Castonguay, L. G., Hayes, J. A., & Janis, R. A. (2025). Teletherapy Post-COVID-19: Comparisons With In-Person Client Characteristics and Service Utilization in Routine Practice. Journal of Clinical Psychology. https://doi.org/10.1002/jclp.70039 

 

Appendix 

Data Analysis 

We used one-to-one nearest neighbor propensity score matching to create a sample of 17,416 clients receiving either IP or TH psychotherapy in routine practice at 138 university counseling centers in the United States. Within each treatment modality, we contructed stacked ensemble models combining 11 machine learning algorithms: LASSO regression, ridge regression, elastic net, random forest, gradient boosting, nearest neighbors, boosted tree, support vector machine, multivariate adaptive regression splines, cubist regression, and classification trees. These were trained to predict pre- to post-treatment symptom change. The differences between predicted outcomes under actual and counterfactual treatment modality were used to estimate individual treatment effects (ITEs) and to determine the optimal treatment modality for each client. A subsequent penalized regression model examined client moderators of treatment effect, which included 32 pre-treatment variables comprising baseline symptoms, treatment history, suicidal and homicidal risk, and demographic characteristics. Overall predicted improvement from assigning clients to their optimal treatment mode was then evaluated.  

Results 

The stacked models showed modest predictive performance (R² = .164 for in-person and .187 for telehealth). Predicted ITEs ranged from -0.45 to 0.36 with a mean of −0.01 (SD = 0.06), with slightly more clients predicted to benefit from TH treatment than IP (58% vs 42%). This indicated minimal differences in symptom change between modalities. The second-stage model explained very little variance in ITEs (R² = 0.01), with small predictor coefficients ranging from -0.01 (Asian/Asian American identity) to 0.03 (baseline Academic Distress). All coefficients were less than half the standard deviation of the ITE distribution. The estimated optimal treatment modality gain was 0.03 points, corresponding to 0.12 standard deviation units. 

 

 Full List of Predictor Variables 

Age 

Financial Stress (current) 

Race/Ethnicity 

  • African American/Black 
  • American Indian or Alaska Native 
  • Asian/Asian American 
  • Hispanic/Latina/o/e 
  • Multi-racial 
  • Native Hawaiian or Pacific Islander 
  • Self-Identify 

Gender Identity 

  • Cisgender man 
  • Cisgender woman 
  • Transgender/non-binary 

Sexual Orientation 

  • Asexual 
  • Bisexual 
  • Gay 
  • Lesbian 
  • Pansexual 
  • Queer 
  • Questioning 

Treatment History 

  • Prior therapy 
  • Prior psychiatric medication 
  • Prior psychiatric hospitalization 

Risk-Related Variables 

  • Prior non-suicidal self-injury 
  • Prior suicide attempt 
  • Intentionally caused serious physical injury to someone else (lifetime) 
  • Seriously considered suicide (lifetime) 

CCAPS-34 Subscales (first administration) 

  • Depression 
  • Generalized Anxiety 
  • Social Anxiety 
  • Academic Distress 
  • Eating Concerns 
  • Anger/Frustration 
  • Alcohol Use 
Published November 25, 2025

Share this post: