New Method of Depressive Symptom Analysis From The Netherlands

Out on PubMed, from investigators in the Netherlands, is this study:

Dynamic time warp analysis of individual symptom trajectories in depressed patients treated with electroconvulsive therapy.

Booij MM, van Noorden MS, van Vliet IM, Ottenheim NR, van der Wee NJA, Van Hemert AM, Giltay EJ.J Affect Disord. 2021 Jul 2;293:435-443. doi: 10.1016/j.jad.2021.06.068. Online ahead of print.PMID: 34252687

The abstract is copied below:

Background: Although electroconvulsive therapy (ECT) effectively improves severity scores of depression, its effects on its individual symptoms has scarcely been studied. We aimed to study which depressive symptom trajectories dynamically cluster together in individuals as well as groups of patients during ECT using Dynamic Time Warp (DTW) analysis.

Methods: We analysed the standardized weekly scores on the 25-item abbreviated version of the Comprehensive Psychopathological Rating Scale (CPRS) in depressed patients before and during their first six weeks of ECT treatment. DTW analysis was used to analyse the (dis)similarity of time series of items scores at the patient level (300 'DTW distances' per patient) as well as on the group level. Hierarchical cluster, network, and Distatis analyses yielded symptom dimensions.

Results: We included 133 patients, 64.7% female, with an average age of 60.4 years (SD 15.1). Individual DTW distance matrices and networks revealed marked differences in hierarchical and network clusters among patients. Based on cluster analyses of the aggregated matrices, four symptom clusters emerged. In patients who reached remission, the average DTW distance between their symptoms was significantly smaller than non-remitters, reflecting denser symptom networks in remitters than non-remitters (p=0.04).

Limitations: The assessments were done only weekly during the first six weeks of ECT treatment. The use of individual items of the abbreviated CPRS may have led to measurement error as well as floor and ceiling effects.

Conclusion: DTW offers an efficient new approach to analyse symptom trajectories within individuals as well as groups of patients, aiding personalized medicine of psychopathology.

The pdf is here.

And rom the statistical analysis section:

We used DTW to calculate distances between each pair of symptoms within each patient.(Berndt and Clifford, 1994) DTW is an approximate pattern detection algorithm that measures dissimilarity between two temporal sequences. It uses a dynamic (i.e., stretching and compressing) M.M. Booij et al. programming approach to minimize a predefined distance measure, in order for the two time-series to become optimally aligned through a warping path. The ‘optimal’ alignment minimizes the sum of distances between aligned elements. The DTW function computes the alignment with a symmetric continuity constraint, implying that arbitrary time compressions and expansions are allowed, and that all elements must be matched. We used the step pattern ‘symmetric P0’ and the global constraint of a ‘Sakoe-Chiba’ window band of 1 around main diagonal (Fig. 1).(Sakoe and Chiba, 1978) This implied that only changes in symptom scores that were maximally 1 time points away (plus or minus 1 weeks) were used within the chosen warping window. As a result, items with the best alignment, having a more similar slope and other dynamics (i.e., changes that co-vary over time) resulted in the smallest distance. All item scores were standardized (i.e., z-scores) before the DTW analysis, as some (e.g., “11. Compulsive thoughts”) tended to score lower on average than other items (e.g., “4. Anhedonia”). This prevents clustering together of symptoms that tend to have a similar average score across patients. Only to ease the interpretation of the DTW technique in Fig. 1, we used unstandardized item scores to visualize the distance calculation in this individual patient.

And from the Discussion:

The results of our analyses in 133 depressive patients treated with ECT show that the trajectories and clusters of all individual depressive symptoms vary substantially between patients during treatment. These differences may unmask clinical information which is typically not visible when only the sum scores of symptom severity scales are presented.(Fried and Nesse, 2015) We were able to cluster the individual CPRS symptoms at the individual level, and also at the group level, which revealed four symptom clusters on basis of their similar dynamics over time. This clustering was robust as shown by the similar findings in two samples after a random sample split of our patients, and the Distatis analysis and the average distance network also revealed similar symptom dimensions in the total sample. We assume that the clustering of symptoms is the result of (potential causal) interactions between symptoms especially with in each of the symptom clusters.

And from the Conclusion:

We conclude that the clustering of symptom trajectories in severely depressed patients showed large variability. This per-patient and per- item variability may reflect the heterogeneity of depression and the delicate (causal) relationships between symptoms. We recommend integrating such DTW analysis on the dynamic symptom network to the personalized medicine approach (Fried et al., 2017) in observational as well as intervention studies. Our findings need to be replicated in larger study samples, with smaller time intervals (such as in Ecological momentary assessment (EMA) or ESM), and over longer observation periods which should also include the dynamics of symptoms before the start of treatment.

To be perfectly honest, I cannot tell whether this paper is genius or gobbledygook, or something in between. Since it is published in the Journal of Affective Disorders, I will give it the benefit of the doubt and call it semi-genius. It uses dastardly complicated statistics and mathematics to analyse individual depressive symptoms and symptom cluster trajectories in individual patients and groups; the authors nicely characterize depression as a "complex dynamic system."

There are no content findings of note, rather only the process finding that this method of analysis may be helpful to predict responders/non-responders to ECT in the future.

It's a brave new world and I may be a Luddite; perhaps that is why my brain is still hurting after reading this paper. But on the bright side, it did make me think of the "Rocky Horror Picture Show."

I hope some blog followers will read every word of this paper (~40 minutes) and let us all know your opinions, thanks.