Skip to main content
Full access
Articles
Published Online: 30 January 2019

Proof of Concept for an Adaptive Treatment Strategy to Prevent Failures in Internet-Delivered CBT: A Single-Blind Randomized Clinical Trial With Insomnia Patients

Abstract

Objective:

This study aimed to demonstrate proof of concept for an adaptive treatment strategy in Internet-delivered cognitive-behavioral therapy (ICBT), where risk of treatment failure is assessed early in treatment and treatment for at-risk patients is adapted to prevent treatment failure.

Methods:

A semiautomated algorithm assessed risk of treatment failure early in treatment in 251 patients undergoing ICBT for insomnia with therapist guidance. At-risk patients were randomly assigned to continue standard ICBT or to receive adapted ICBT. The primary outcome was self-rated insomnia symptoms using the Insomnia Severity Index in a linear mixed-effects model. The main secondary outcome was treatment failure (having neither responded nor remitted at the posttreatment assessment).

Results:

A total of 102 patients were classified as at risk and randomly assigned to receive adapted ICBT (N=51) or standard ICBT (N=51); 149 patients were classified as not at risk. Patients not at risk had significantly greater score reductions on the Insomnia Severity Index than at-risk patients given standard ICBT. Adapted ICBT for at-risk patients was significantly more successful in reducing symptoms compared with standard ICBT, and it decreased the risk of failing treatment (odds ratio=0.33). At-risk patients receiving adapted ICBT were not more likely to experience treatment failure than those not at risk (odds ratio=0.51), though they were less likely to experience remission. Adapted treatment required, on average, 14 more minutes of therapist-patient time per remaining week.

Conclusions:

An adaptive treatment strategy can increase treatment effects for at-risk patients and reduce the number of failed treatments. Future studies should improve accuracy in classification algorithms and identify key factors that boost the effect of adapted treatments.
Personalized medicine holds great promise in improving the quality of treatments for mental health conditions (1), whereas the concept of stepped care has the potential to make these interventions more available and cost efficient (2). With personalized care, treatment outcomes are improved by giving individual patients a treatment that is precisely right for them (3). In stepped care, on the other hand, all or most patients are given the same, least intensive, most readily available treatment. Those who require additional treatment are then moved up a step to intensified treatment. Personalized medicine is usually thought of as a way to match patients to the optimal intervention before initiating treatment (3, 4) because no intervention is likely to be ideal for all. If successful, personalized medicine would overcome a shortcoming of stepped care, where many patients have to go through a number of steps before reaching a treatment that is effective for them and consequently be exposed to treatment failure. In fact, a large evaluation of the Improving Access to Psychological Therapies project, an implemented stepped-care system, showed that moving up through steps was the strongest predictor of recovery (5). This may indicate that low-level steps are inappropriate starting points for many patients. Starting, and working through, ineffective treatments may result in prolonged suffering and potentially unnecessary costs.
A wide range of studies have identified baseline predictors of treatment outcome in psychological treatment, but with mixed findings. To justify patient-treatment matching, predictions would need to be very strong. Although there have been ambitious attempts, such as Project MATCH for alcohol cessation (6), no study thus far has developed and confirmed the usefulness of tools that prospectively match patients to treatments.
However, the concept of personalized care need not be restricted to patient-treatment matching before treatment initiation; it could also include adjusting treatment on the basis of cues of treatment nonresponse. For instance, adjusting the dosage of a selective serotonin reuptake inhibitor based on detection of cytochrome genetic polymorphism or switching to another compound could be framed as ways to personalize treatment. In relation to psychological treatment, studies have shown that clinical data collected during treatment are more reliable in predicting outcome than baseline data (7, 8). Lambert and colleagues have repeatedly shown that progress monitoring, based on structured patient self-report and feedback tools for therapists, improves outcomes, especially for patients who are predicted not to respond to a treatment, or to deteriorate (9). Beyond providing the therapist with a prediction, a plan for how the therapist should act on the prediction is thought to improve the effect further (9), but this has never been tested experimentally.
In this study, we aimed to demonstrate a proof of concept and develop and test an adaptive treatment strategy by using baseline data and data from the first few treatment weeks to predict final outcome, by presenting this prediction to the therapist in a comprehensible way, and, in cases where an unsuccessful outcome is predicted, by providing a method to intensify and personalize the ongoing treatment to adapt it to the needs of the patient. If successful, this could help transform stepped care into more of an accelerated care, where entry-level treatments can be adaptive, thus avoiding the need to wait for an initial treatment attempt to fail before intensifying care.
To increase the probability of a successful demonstration, we made two choices. First, the context of therapist-supported, Internet-delivered cognitive-behavioral therapy (ICBT) was chosen because it is empirically supported, allows for easily collected data, and presents several possible adaptations regarding intensity and delivery (10). This treatment format has also been implicated as a key first-step intervention in stepped-care models (1113). Second, we chose ICBT for patients with insomnia because it is a well-delimited clinical area with well-defined therapeutic techniques facilitating the assessment of problems related to progress and adherence, while also having strong evidence (14, 15) but room for overall improvement (16).
We hypothesized that:
1.
Treatment failure can be predicted by using an algorithm-based individual outcome prediction tool early in treatment (i.e., using the tool to classify patients as at risk or not at risk and comparing them), and patients at risk who receive standard treatment will benefit less from treatment than patients not at risk.
2.
A structured adaptation of the treatment is beneficial for patients predicted to experience treatment failure; that is, patients at risk who receive adapted treatment benefit as much as patients not at risk and benefit more than patients at risk who receive standard treatment.

Methods

Design

The study was a single-blind randomized controlled trial with patients undergoing ICBT for insomnia. During week 4 out of 9, patients were classified as Red (at risk of experiencing treatment failure) or Green (not at risk of experiencing treatment failure), and patients in the Red group were randomized to adapted treatment (Red-Adapted), aimed at preventing failure, or to continue with standard care (Red-Standard). The study was approved by the regional ethical review board in Stockholm.

Recruitment and Inclusion Criteria

Participants living in Stockholm County were recruited via media and the Internet Psychiatry Clinic’s web site (www.internetpsykiatri.se), where they filled out screening questionnaires and then visited the clinic for an assessment with a psychiatrist, resident physician, and/or clinical psychologist. The assessment included the Mini International Neuropsychiatric Interview (17) for DSM-IV-TR and a suicide risk assessment. All patients provided written informed consent at intake.
Inclusion criteria were the following: 18 years or older; insomnia diagnosis; score >10 on the Insomnia Severity Index (18); stable or no antidepressant medication for at least 2 months; no diseases, disorders, or substance abuse that required other, immediate attention (e.g., severe depression or suicidality) or that was considered a contraindication for the treatment in the present trial (e.g., sleep apnea, bipolar disorder, complex multimorbidity deemed to hinder treatment); no night-shift work; available time for treatment and acceptance of its format; proficient in Swedish; and no ongoing CBT. Patients with depression were excluded and recruited by another study. Use of sleep medication was not an exclusion criterion.

Measures

Patients filled out questionnaires at screening, pretreatment, and posttreatment assessments, as well as weekly during treatment. At the posttreatment assessment, patients who did not fill out online questionnaires were interviewed by telephone.

Primary outcome.

The primary outcome measure was change over time in patient-rated insomnia severity as measured with the Insomnia Severity Index, a seven-item scale ranging from 0 to 28 points (19). The scale is reliable and sensitive to change (18, 19).

Secondary outcomes.

Treatment failure was defined as neither responding nor remitting. Response was defined as a decrease in the Insomnia Severity Index score of at least 8 points (18), and remission was defined as a score <8 on the Insomnia Severity Index (18). We defined deterioration as an increase of ≥3 points in Insomnia Severity Index score, in accordance with Wise (20). A complementary binary outcome was a 50% reduction in symptom ratings (21).

Other measures.

The amount of time spent by therapists writing to patients and talking to patients on the telephone, the number of messages sent and telephone calls made, and the number of modules completed were measured.
To control for differential use of sleep medication and other treatments during the study period, current use and changes from baseline were assessed after treatment via questionnaires. Patients were not required to keep their medication stable or to make any changes they did not wish to make, but stabilizing and tapering sleep medication were encouraged, and a structured way of doing this was an optional part of the treatment.
The eight-item Client Satisfaction Questionnaire was used at the posttreatment assessment. Scores range from 8 to 32; scores of 8–13 indicate poor satisfaction, 14–19 fair satisfaction, 20–25 good satisfaction, and 26–32 excellent satisfaction (22). Patients were also asked an open-ended question about adverse events or negative effects.

The Adaptive Treatment Strategy

The standard ICBT treatment.

The treatment was a 9-week therapist-supported ICBT program for insomnia, highly similar to a previously evaluated program showing large and lasting symptom reductions (23), comparable to face-to-face therapy (24). Treatment was based on a well-established manual with sleep restriction and stimulus control as the main therapeutic components, complemented with other techniques as presented in Appendix A in the online supplement. There were 25 therapists: 13 licensed clinical psychologists and 12 final-year students in a 5-year clinical psychology program that includes 18 months of supervised CBT practice. Those who were licensed had been so for a median of 1.5 years (range=0–7). All therapists were novices regarding ICBT for insomnia. Therapists were given a half-day education on ICBT for insomnia and the study protocol. New therapists initially also had the content of their text messages with patients checked by and discussed with a supervisor. Therapists were instructed to spend a maximum of 15 minutes per week guiding patients via text messages within the platform. Weekly group supervision by K.B. or S.J. was held to ensure that therapists adhered to the study protocol. The allocation of licensed and nonlicensed therapists after randomization was even across groups (55% licensed therapists per Red group). To control for any influence of therapist competence, sensitivity analyses of the main statistical models, with therapist competence as a covariate, were performed. All patients were blind to classification and randomization.

The individual outcome prediction algorithm.

The individual outcome prediction was carried out during week 4 in treatment and was preceded by a specific questionnaire. Patients had previously been informed of this additional assessment but not of the classification and randomization. A description and a working copy of the tool are presented in Appendix B and Appendix C in the online supplement. Patients were classified as Green if the prediction algorithm suggested that they would benefit from treatment and Red if the prediction algorithm suggested that they might not benefit. Patients who had already discontinued participation in the study or were unable to continue were classified as dropouts in the first step of the algorithm.

Randomization.

Immediately after classification, patients in the Red group were randomly assigned to receive either adapted treatment (Red-Adapted) or standard treatment (Red-Standard), which meant continuing the treatment just as before classification. Randomization was carried out using sealed and numbered envelopes. The randomization and sealing of the envelopes were done in blocks of random size and order and were carried out by an independent person. Opening the envelopes and documenting the outcome of randomization were done consecutively by an independent person. The group allocation is presented in Figure 1.
FIGURE 1. CONSORT diagram of participants in a study of an adaptive treatment strategy to prevent treatment failures in Internet-delivered CBT for insomniaa
a Data include some specifications to the exclusion criteria for individual patients frequently reported by assessors. However, because reporting a specification was not mandatory, the reported frequencies are minimum rather than absolute values. CBT=cognitive-behavioral therapy; ISI=Insomnia Severity Index.

The adapted treatment.

Patients in the Red-Adapted group were contacted by their therapist for a semistructured telephone interview directly after randomization. The interview took about 30 minutes and aimed to assess the possible causes for the patient’s difficulties with treatment and to examine potential alterations to the treatment format that might help the patient.
The therapist was allowed to suggest and agree to adaptations concerning the treatment format, the intensity and modality of contact, and the level of therapist directiveness and patient autonomy. Adaptation would most commonly consist of telephone support, printed materials via regular mail, and/or increased text-message reminders. The adaptation was to intensify the treatment rather than change its content; thus, the focus would still be on insomnia, and the therapist was not allowed to suggest, or agree to, any adaptation or added content pertaining to a comorbid problem or disorder (e.g., worry, anxiety, relationship problems). There was no strict upper time limit for therapists to spend on patients in adapted treatment as long as that time was kept within working hours and all other patients were treated according to their respective protocols without compromise. The supervisors would follow the extra support given and ascertain that the level of support was within reason. Several fictive cases of potential patients in adapted treatment, with the proposed changes to treatment and therapist support, are presented in Appendix D in the online supplement.

Statistical Analyses

Analyses were done according to the intention-to-treat principle, where patients who received a Red or Green designation were included in the study.
A longitudinal multilevel model with random slopes and random intercepts was used for the main analysis to provide a more accurate representation of reality than single-level linear models; this model also uses all available data points and therefore minimizes information loss due to missing data (25). The time points used were the pretreatment assessment, weekly measures, and the posttreatment assessment.
To test our first hypothesis, we used a latent growth curve model with random intercepts and slopes covering all time points for the Green and Red-Standard groups.
To test our second hypothesis, we used a two-period latent growth curve model with random intercepts and slopes, where one period is that from pretreatment assessment to classification, and the other is the period from classification to posttreatment assessment. This approach isolates the time frame where we expect to see a difference in trajectories between the Red groups.
For baseline comparisons of sample characteristics and comparisons of secondary outcomes, two-way analyses of variance and two-tailed t tests, chi-square contingency tables, and odds ratios were used when applicable. Statistical analyses were carried out using SPSS, version 25 (IBM, Armonk, N.Y.).

Results

A total of 251 patients were included and subjected to the classification procedure, in which 149 were classified as Green and 102 as Red.

Missing Data

Three of the six missing online posttreatment Insomnia Severity Index scores could be replaced with interview versions of the Insomnia Severity Index, in accordance with Hedman et al. (26).
Of 10 measurement points, 71% of participants provided data for all 10 points, and 91% provided data for eight or more points. Attrition at posttreatment assessment was low, at 1.2%. No participant provided fewer than two data points, and therefore all participants are included in the estimation of all coefficients in the primary outcome analyses (25).
Missing data analysis for nonignorable missing data revealed no variables with significant correlations to both outcome and to being absent.

Patient Baseline Characteristics

Baseline characteristics of the different classified and randomized groups are summarized in Table 1. Having completed postsecondary education was significantly less common in the Red-Standard group compared with the Red-Adapted group (χ2=5.67, df=1, p=0.017). As a sensitivity analysis, the two linear mixed-model analyses were repeated with postsecondary education as a covariate, which did not change the results.
TABLE 1. Baseline characteristics of participants in a study of an adaptive treatment strategy to prevent treatment failures in Internet-delivered cognitive-behavioral therapy for insomnia
CharacteristicGreen Group (N=149)Red-Standard Group (N=51)Red-Adapted Group (N=51)
 MeanSDMeanSDMeanSD
Age (years)47.813.943.414.346.212.5
 N%N%N%
Female9865.83670.63976.5
With children9966.43262.73466.7
Married or de facto married9261.73262.73058.8
Education achieved
 Primary42.759.835.9
 Secondary4932.92549.01529.4
 Postsecondary9664.42141.23364.7
Occupational status
 Currently working or student12382.64180.44078.4
 Full-time occupation9966.43058.83160.8
Use of sleep medication9765.12650.92752.9
Comorbid axis I disorder (excluding depression)2818.81223.51019.6
 MeanSDMeanSDMeanSD
Insomnia Severity Index score16.84.318.33.719.04.2

Primary Outcomes

The means and standard deviations of scores on the Insomnia Severity Index across groups are presented in Table 2, and score trajectories across all measurement points are shown in Figure 2. There were significant main effects of time for both models (p<0.001), indicating that all three groups improved from pretreatment to posttreatment assessments. There was also a significant effect of group at the intercept (p<0.001), indicating that patients in the Red group had higher pretreatment scores on average than those in the Green group.
TABLE 2. Participant scores on the Insomnia Severity Indexa
Participant GroupPretreatmentClassification (week 3)Posttreatment
MeanSDMeanSDMeanSD
Green16.84.311.74.57.54.5
Red-Adapted19.04.216.64.110.74.9
Red-Standard18.33.715.53.712.44.2
a
Patients in the Green group were classified as likely to have a good outcome and received the standard treatment. Patients in the Red-Adapted group were classified as at risk of failing to respond to standard treatment and then received an adapted treatment. Patients in the Red-Standard group were classified as at risk of failing to respond to standard treatment and then still received the standard treatment.
FIGURE 2. Observed trajectories of insomnia severity over time for the three patient groupsa
a Patients in the Green group (N=149) were classified as likely to have a good outcome and received the standard treatment. Patients in the Red-Adapted group (N=51) were classified as at risk of failing to respond to standard treatment and then received an adapted treatment. Patients in the Red-Standard group (N=51) were classified as at risk of failing to respond to standard treatment and then still received the standard treatment.

Hypothesis 1: treatment failure can be predicted.

There was a significant group-by-time interaction indicating larger reductions over time in symptom ratings for the Green group compared with the Red-Standard group (p<0.001), supporting our first hypothesis.

Hypothesis 2: treatment adaptation is beneficial.

Before classification, both Red groups had significantly worse treatment effects compared with the Green group (p<0.001 for both Red groups) but were not significantly different from each other (p=0.775). After classification, the Red-Adapted group improved significantly more (p<0.001) than the Red-Standard group and significantly more than the Green group (p<0.001). This result indicates that the rate of improvement accelerated when treatment was adapted, in support of our second hypothesis. After classification, Green and Red-Standard patients did not differ significantly in change over time (p=0.214).

Treatment Failure, Response, Remission, and Deterioration

Categorical outcomes are summarized in Table 3. Treatment failure was significantly more common among patients in the Red-Standard group than those in the Green and Red-Adapted groups, who did not differ significantly from each other. Remission was significantly more common in the Green group than the other groups and was not statistically different between Red groups. Responder status and reductions ≥50% in symptom ratings were not significantly different between the Green and Red-Adapted groups, and patients in the Red-Adapted group had significantly higher response rates and ≥50% reductions in symptom ratings compared with those in the Red-Standard group. Deterioration was too rare in all groups to produce meaningful statistical comparisons.
TABLE 3. Categorical treatment outcomes (observed data) among participants in a study of an adaptive treatment strategy to prevent treatment failures in Internet-delivered cognitive-behavioral therapy for insomniaa
 Green Group (N=149)Red-Adapted Group (N=49)Red-Standard Group (N=50)Red-Adapted Compared With Red-StandardGreen Compared With Red-StandardGreen Compared With Red-Adapted
OutcomeN%N%N%Odds RatioOdds RatioOdds Ratio
Failure3422.81836.73264.00.327**0.166***0.509
Response9865.82956.91835.32.58*3.42***1.33
Remission8154.41325.5611.82.658.74***3.29**
>50% symptom reduction9463.1247.11325.52.73*4.86***1.78
Deteriorated21.300.012.00.9800.6671.01
a
Response was defined as a decrease of ≥8 points in Insomnia Severity Index score. Remission was defined as a score <8 on the Insomnia Severity Index at the posttreatment assessment. Treatment failure was defined as neither responding nor remitting. Patients in the Green group were classified as likely to have a good outcome and received the standard treatment. Patients in the Red-Adapted group were classified as at risk of failing to respond to standard treatment and then received an adapted treatment. Patients in the Red-Standard group were classified as at risk of failing to respond to standard treatment and then still received the standard treatment.
*
p<0.05.
**
p<0.01.
***
p<0.001.

Therapist Adherence to the Adaptive Treatment Strategy

Therapist adherence was good. Therapists spent, on average, 110.9 minutes per patient writing messages and talking on the telephone with patients in standard treatment overall, compared with 179.6 minutes per patient in adapted treatment (t=7.7, df=249, p<0.001); this equates to a little more than 1 extra hour per treatment, or about 14 extra minutes per week, for the remaining 5 weeks of treatment. The same pattern was observed in the number of message or telephone contacts (17.3 in standard treatment compared with 19.8 in adapted treatment; t=3.4, df=249, p=0.001). When comparing the Green with the Red-Standard group, the total time spent by therapists was lower for the Red-Standard group (24.3 minutes less in total, or 2.7 minutes less per week; t=3.1, df=198, p=0.003). However, that difference disappears when looking at time spent per module (t=1.1, df=198, p=0.269) because the Red-Standard group completed fewer modules (t=8.6, df=198, p<0.001) and therefore elicited fewer responses and less feedback from therapists. There was no significant effect of therapist licensure on change in scores on the Insomnia Severity Index or in any of the categorical outcomes. As a sensitivity analysis, the two linear mixed-model analyses were repeated with assignment to a licensed therapist or not as a covariate, and the results did not change.

Medication Changes and Use of Other Treatments

More patients in the Green group than in the Red groups stopped taking sleep medication (Green group compared with Red-Adapted group, χ2=6.76, df=2, p=0.03; Green group compared with Red-Standard group, χ2=11.45, df=2, p=0.003). Eight patients in the Red-Adapted group (16%) and six in the Red-Standard group (12%) had stopped taking sleep medication at the posttreatment assessment, which was not a statistically significant difference (χ2=0.619, df=2, p=0.73). There were no statistically significant differences between groups in terms of starting sleep medication or increasing or decreasing its dosage. A significantly greater number of patients in the Red-Standard group reported having started some other intervention (CBT, medication, or other) to reduce sleep problems during the study period, compared with patients in the Red-Adapted and Green groups: six (13.3%), two (4.1%), and five (3.4%), respectively (Fisher’s exact p=0.039).

Treatment Satisfaction and Adverse Events

Client Satisfaction Questionnaire scores were good to excellent (22) in all three groups (Red-Standard, mean=24.1, SD=4.1; Red-Adapted, mean=24.7, SD=4.8; Green, mean=27.1, SD=4.1) but were significantly higher among patients in the Green group compared with the other two groups (F=11.63, df=2, 24, p<0.001).
No serious adverse events were reported. Reporting of any adverse event or negative effect in response to an open-ended question at the posttreatment assessment was not significantly different between any groups, although it was slightly less common among patients in the Red-Adapted group (24%) than among those in the Red-Standard (31%) or Green (30%) groups. The most common adverse events were sleep getting worse in conjunction with starting sleep restriction (41% of adverse events, or 12% of the full sample) and an increase in physical symptoms (23% of adverse events, or 7% of the full sample), most commonly a subjective experience of increased susceptibility to getting a cold or the flu while feeling sleep deprived.

Discussion

Our aim in this study was to investigate the usefulness of an adaptive treatment strategy using an individual outcome prediction algorithm and adapting treatment for patients at risk of treatment failure. As hypothesized, patients classified as at risk of experiencing treatment failure who did not receive adapted treatment (Red-Standard) had significantly less symptom reduction than patients identified as probable successes (Green). Furthermore, patients classified as at risk of experiencing treatment failure who received the adapted treatment (Red-Adapted) had significantly greater symptom reduction than at-risk patients who received standard treatment (Red-Standard).
Although at-risk patients who received adapted treatment improved more than those who received standard treatment, they did not do quite as well as the patients who were not at risk. This could be because at-risk patients had markedly higher pretreatment symptom levels, but it could also indicate that some at-risk patients are not adequately helped by this type of adaptation to treatment. However, once adapted treatment was given, at-risk patients showed a steeper decrease in symptom rating than not-at-risk patients, so they might have caught up if treatment had been extended. Importantly, the adapted treatment offered in this trial does not seem to be very time consuming. Therapists spent, on average, only about 14 more minutes per remaining week on these patients, and that time includes the 30-minute adaptation assessment. In addition, the adaptations were limited to format and intensity only, and no novel interventions were added.
Our findings are in line with those of Lambert (9) in that structured monitoring and identification of patients who are not progressing well, coupled with active intervention from therapists, can improve outcomes in psychotherapy in general. In the present trial, the treatment adaptations and therapists’ behaviors were uniquely monitored and restricted to certain types of adaptation only, providing important evidence of the utility of relatively simple adaptations.
One key difference is in the rates of deterioration. Hannan et al. (27) observed about 7% of patients leaving treatment as deteriorated, and Shimokawa et al. (28) found 20% deterioration without support tools, which was reduced to 11% with support tools. In our study, 1.2% of patients deteriorated without adapted treatment and 0% did so with adaptation. However, deterioration rates in ICBT could be lower than in psychotherapy in general (29), and insomnia is often chronic and stable, possibly making patients unlikely to deteriorate.
The number of reported adverse events is higher than what is usually reported in psychotherapy research. However, adverse events have most often not been explicitly measured and reported in previous research, as noted by Rozental et al. (30), suggesting an underestimation in previous findings. Also, regarding CBT for insomnia, the number of reported adverse events in this study does in fact match previous findings relatively well (e.g., 31, 32). The intervention itself (sleep restriction) is associated with a number of specific, acute side effects that are distressing but at the same time passing and expected, or even considered hypothetical mechanisms (31, 32).
Important strengths of this trial are the random assignment to adapted treatment for at-risk patients and the patients’ blind to both assignment and classification, especially because the primary outcome is patient rated. Moreover, there was very little attrition and missing data.
One limitation is that therapists were aware that at-risk patients were predicted to experience treatment failure, which could affect therapist performance (causing them to give up or to put in extra effort). However, this effect is less likely in the context of this study’s Internet-delivered treatment, where much of the treatment is standardized and therapists were supervised in terms of time spent and what to do with patients in standard treatment (i.e., not to undertreat them).
Another limitation is that time spent mailing materials to patients in the adapted treatment group or trying, but failing, to reach someone on the telephone was not measured. At-risk patients in standard care received less time overall probably because they were inactive, illustrated by the fact that time spent per message did not differ. Overall, our data indicate that the therapists followed the adaptive treatment strategy protocol for all groups.
The use of ICBT for insomnia reduces the generalizability of these findings, as it is possible that other treatment formats or other target conditions would respond differently to the adapted treatment. On the other hand, the standardized ICBT made the comparison less prone to confounding by other factors.
The individual outcome prediction tool was not validated beforehand, but instead validation and trial of the adapted treatment were parallel. Had the algorithm been evaluated beforehand, it might have been more accurate and thus produced clearer results. Although the results of this trial are already quite clear, further studies are still needed to tune and automate such algorithms to maximize their utility.
Another limitation is the lack of follow-up data. Further studies are needed to determine the long-term effects of the adapted treatment and the progression of at-risk patients who do not receive adapted treatment over time. The aim of the present study, however, was to investigate immediate effects of using an adaptive treatment strategy.
Future studies should address how the proposed model works in regular care, where it may be difficult to provide adapted treatment only to at-risk patients or to address only one condition. However, focusing on only one condition or problem may be beneficial in itself, at least in insomnia, as evidence suggests that providing a focused insomnia intervention can be beneficial even for patients with comorbidities (33, 34). In a broader health care delivery context, interventions utilizing adaptive treatment strategies could, together with the prospective matching of patients to treatments, transform the stepped-care paradigm into accelerated care, in order to shorten the time of suffering by minimizing time spent in low-intensity but nonhelpful interventions. In psychotherapy, where patient-treatment matching is still highly elusive, adaptive treatment strategies may be easier to create and implement in a useful way. For most patients, it may be more useful and accurate to determine early on that a treatment is not working, rather than trying to predict treatment failure before the treatment starts. In some implemented stepped-care systems today, it is possible that patients are already being moved to the next step before an ongoing treatment course is formally over if nonresponse or deterioration is noticed, or when a patient drops out. To our knowledge, however, moving a patient to the next step has so far never been done on the basis of empirical findings from randomized controlled trials or validated prediction algorithms.
In conclusion, our results suggest that an adaptive treatment strategy can be a feasible way to deliver psychiatric care and that it can increase treatment effects for at-risk patients and reduce the number of failed treatments. The classification algorithm and the adaptation of the treatment should be further developed, but this proof-of-concept study indicates that adaptive treatment strategies could be an important step in moving from stepped care into accelerated care.

Acknowledgments

The authors thank the clinical and administrative staff at the Internet Psychiatry Clinic in Stockholm as well as all the master’s-level students who helped conduct this trial, and especially research nurse Monica Hellberg. Most of all, the authors thank all of the participating patients.

Footnote

ClinicalTrials.gov identifier: NCT01663844.

Supplementary Material

File (appi.ajp.2018.18060699.ds001_appendicesa-b-d.pdf)
File (appi.ajp.2018.18060699.ds002_appendixc.xlsx)

References

1.
Fernandes BS, Williams LM, Steiner J, et al: The new field of “precision psychiatry”. BMC Med 2017; 15:80
2.
Bower P, Gilbody S: Stepped care in psychological therapies: access, effectiveness and efficiency: narrative literature review. Br J Psychiatr, 2005, pp 186
3.
Goldberger JJ, Buxton AE: Personalized medicine vs guideline-based medicine. JAMA 2013; 309:2559–2560
4.
Hallgren KA, Bauer AM, Atkins DC: Digital technology and clinical decision making in depression treatment: current findings and future opportunities. Depress Anxiety 2017; 34:494–501
5.
Gyani A, Shafran R, Layard R, et al: Enhancing recovery rates: lessons from year one of IAPT. Behav Res Ther 2013; 51:597–606
6.
Matching Alcoholism Treatments to Client Heterogeneity: Project MATCH posttreatment drinking outcomes. J Stud Alcohol 1997; 58:7–29
7.
Schibbye P, Ghaderi A, Ljótsson B, et al: Using early change to predict outcome in cognitive behaviour therapy: exploring timeframe, calculation method, and differences of disorder-specific versus general measures. PLoS One 2014; 9:e100614
8.
Lutz W, Arndt A, Rubel J, et al: Defining and predicting patterns of early response in a web-based intervention for depression. J Med Internet Res 2017; 19:e206
9.
Lambert MJ: Progress feedback and the OQ-system: the past and the future. Psychotherapy (Chic) 2015; 52:381–390
10.
Espie CA, Hames P, McKinstry B: Use of the Internet and mobile media for delivery of cognitive behavioral insomnia therapy. Sleep Med Clin 2013; 8:407–419
11.
Andrews G, Titov N: Changing the face of mental health care through needs-based planning. Aust Health Rev 2007; 31(Suppl 1):S122–S128
12.
Andrews G: World Health Organization Collaborating Centre for Classification in Mental Health: Tolkien II: a needs-based, costed, stepped-care model for mental health services: recommendations, executive summaries, clinical pathways, treatment flowcharts, costing structures. Sydney, N.S.W., 2007
13.
National Collaborating Centre for Mental Health: National Institute for Health and Clinical Excellence: Guidance, in Depression: The Treatment and Management of Depression in Adults (Updated Edition). Leicester, UK, British Psychological Society, 2010
14.
Mitchell MD, Gehrman P, Perlis M, et al: Comparative effectiveness of cognitive behavioral therapy for insomnia: a systematic review. BMC Fam Pract 2012; 13:40
15.
Zachariae R, Lyby MS, Ritterband LM, et al: Efficacy of Internet-delivered cognitive-behavioral therapy for insomnia: a systematic review and meta-analysis of randomized controlled trials. Sleep Med Rev 2016; 30:1–10
16.
Harvey AG, Tang NKY: Cognitive behaviour therapy for primary insomnia: can we rest yet? Sleep Med Rev 2003; 7:237–262
17.
Sheehan DV, Lecrubier Y, Harnett Sheehan K, et al: The validity of the Mini International Neuropsychiatric Interview (MINI) according to the SCID-P and its reliability. Eur Psychiatry 1997; 12:232–241
18.
Morin CM, Belleville G, Bélanger L, et al: The Insomnia Severity Index: psychometric indicators to detect insomnia cases and evaluate treatment response. Sleep (Basel) 2011; 34:601–608
19.
Bastien CH, Vallières A, Morin CM: Validation of the Insomnia Severity Index as an outcome measure for insomnia research. Sleep Med 2001; 2:297–307
20.
Wise EA: Methods for analyzing psychotherapy outcomes: a review of clinical significance, reliable change, and recommendations for future directions. J Pers Assess 2004; 82:50–59
21.
Hiller W, Schindler AC, Lambert MJ: Defining response and remission in psychotherapy research: a comparison of the RCI and the method of percent improvement. Psychother Res 2012; 22:1–11
22.
Smith D, Roche E, O’Loughlin K, et al: Satisfaction with services following voluntary and involuntary admission. J Ment Health 2014; 23:38–45
23.
Kaldo V, Jernelöv S, Blom K, et al: Guided Internet cognitive behavioral therapy for insomnia compared to a control treatment: a randomized trial. Behav Res Ther 2015; 71:90–100
24.
Blom K, Tarkian Tillgren H, Wiklund T, et al: Internet- vs group-delivered cognitive behavior therapy for insomnia: a randomized controlled non-inferiority trial. Behav Res Ther 2015; 70:47–55
25.
Hesser H: Modeling individual differences in randomized experiments using growth models: recommendations for design, statistical analysis and reporting of results of Internet interventions. Internet Interv 2015; 2:110–120
26.
Hedman E, Ljótsson B, Blom K, et al: Telephone versus Internet administration of self-report measures of social anxiety, depressive symptoms, and insomnia: psychometric evaluation of a method to reduce the impact of missing data. J Med Internet Res 2013; 15:e229
27.
Hannan C, Lambert MJ, Harmon C, et al: A lab test and algorithms for identifying clients at risk for treatment failure. J Clin Psychol 2005; 61:155–163
28.
Shimokawa K, Lambert MJ, Smart DW: Enhancing treatment outcome of patients at risk of treatment failure: meta-analytic and mega-analytic review of a psychotherapy quality assurance system. J Consult Clin Psychol 2010; 78:298–311
29.
Rozental A, Magnusson K, Boettcher J, et al: For better or worse: an individual patient data meta-analysis of deterioration among participants receiving Internet-based cognitive behavior therapy. J Consult Clin Psychol 2017; 85:160–177
30.
Rozental A, Andersson G, Boettcher J, et al: Consensus statement on defining and measuring negative effects of Internet interventions. Internet Interv 2014; 1:12–19
31.
Kyle SD, Miller CB, Rogers Z, et al: Sleep restriction therapy for insomnia is associated with reduced objective total sleep time, increased daytime somnolence, and objectively impaired vigilance: implications for the clinical management of insomnia disorder. Sleep (Basel) 2014; 37:229–237
32.
Kyle SD, Morgan K, Spiegelhalder K, et al: No pain, no gain: an exploratory within-subjects mixed-methods evaluation of the patient experience of sleep restriction therapy (SRT) for insomnia. Sleep Med 2011; 12:735–747
33.
Blom K, Jernelöv S, Kraepelien M, et al: Internet treatment addressing either insomnia or depression, for patients with both diagnoses: a randomized trial. Sleep (Basel) 2015; 38:267–277
34.
Jernelöv S, Lekander M, Blom K, et al: Efficacy of a behavioral self-help treatment with or without therapist guidance for co-morbid and primary insomnia: a randomized controlled trial. BMC Psychiatry 2012; 12:5

Information & Authors

Information

Published In

Go to American Journal of Psychiatry
Go to American Journal of Psychiatry
American Journal of Psychiatry
Pages: 315 - 323
PubMed: 30696270

History

Received: 15 June 2018
Revision received: 5 October 2018
Accepted: 29 October 2018
Published online: 30 January 2019
Published in print: April 01, 2019

Keywords

  1. Behavior Therapy
  2. Diagnosis And Classification
  3. Psychotherapy
  4. Sleep
  5. Computers

Authors

Details

Erik Forsell, M.Sc. [email protected]
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Susanna Jernelöv, Ph.D.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Kerstin Blom, Ph.D.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Martin Kraepelien, M.Sc.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Cecilia Svanborg, M.D., Ph.D.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Gerhard Andersson, Ph.D.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Nils Lindefors, M.D., Ph.D.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).
Viktor Kaldo, Ph.D.
Center for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet, and Stockholm Health Care Services, Stockholm County Council, Huddinge Hospital, Stockholm (Forsell, Jernelöv, Blom, Kraepelien, Svanborg, Andersson, Lindefors, Kaldo); Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm (Jernelöv); Department of Behavioral Sciences and Learning, Linköping University, Linköping, Sweden (Andersson); and Department of Psychology, Faculty of Health and Life Sciences, Linnaeus University, Växjö, Sweden (Kaldo).

Notes

Send correspondence to Mr. Forsell ([email protected]).

Author Contributions

The first two authors contributed equally to this article.

Competing Interests

The authors report no financial relationships with commercial interests.

Funding Information

Supported by the Söderström Königska Foundation, the L.J. Boëthius Foundation, the Brother Gadelius Foundation, the Stockholm County Council (funding through the Swedish Medical Training and Research Agreement and infrastructure via the Internet Psychiatry Clinic), the Erling-Persson Family Foundation, and the Swedish Research Council.

Metrics & Citations

Metrics

Citations

Export Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

For more information or tips please see 'Downloading to a citation manager' in the Help menu.

Format
Citation style
Style
Copy to clipboard

View Options

View options

PDF/ePub

View PDF/ePub

Get Access

Login options

Already a subscriber? Access your subscription through your login credentials or your institution for full access to this article.

Personal login Institutional Login Open Athens login
Purchase Options

Purchase this article to access the full text.

PPV Articles - American Journal of Psychiatry

PPV Articles - American Journal of Psychiatry

Not a subscriber?

Subscribe Now / Learn More

PsychiatryOnline subscription options offer access to the DSM-5-TR® library, books, journals, CME, and patient resources. This all-in-one virtual library provides psychiatrists and mental health professionals with key resources for diagnosis, treatment, research, and professional development.

Need more help? PsychiatryOnline Customer Service may be reached by emailing [email protected] or by calling 800-368-5777 (in the U.S.) or 703-907-7322 (outside the U.S.).

Media

Figures

Other

Tables

Share

Share

Share article link

Share