Image source: vecteezy.com

The problem with seeking psychotherapy

This is a personal passion project in which I tackle an issue that I have observed over the years regarding mental health support. In simple terms the issue is this: it is not easy for an average person to choose the right type of psychotherapy for them. There are many reasons for this, and as one attempts to make an informed decision about starting psychotherapy a few of them immediately come into view:

  • There are currently hundreds of approaches to psychotherapy and counselling with a variety of techniques and rationales behind them.
  • The classification and naming conventions are arbitrary and inconsistent, which might be completely unhelpful or even confuse the reader further.
  • The descriptions are lengthy and full of psychological lingo.
  • The information is spread out across different sources and is time consuming to go over.
Because of this it would be easy to either become overwhelmed and give up, or just choose your therapist while ignoring their training and approach used. Either way the end result might not be ideal. In my experience, when asked what approach their therapist is using, most people simply don’t know.

My Solution

I aim to identify the main types of psychotherapy and present them in an infographic format with short descriptions outlying key similarities and differences.

Overview

  1. Gathering the data (Web Scraping)
  2. Identifying different types of therapy and counting how often they are mention in the profiles
  3. Measuring similarities between different modalities based on their descriptions (Natural Language Processing)
  4. Assigning psychotherapy types into groups (based on similarity scores and topic modeling)
  5. Putting it all together and visualising the results

Data

  • Since ‘psychotherapist’ is not a protected title in the UK, anyone can legally call themselves one. To ensure my results represent psychotherapists and counsellors with appropriate training, I chose to use data extracted from the main governing bodies for counselling and psychotherapy in the UK – the British Association for Counselling and Psychotherapy (BACP) and the United Kingdom Council for Psychotherap (UKCP). Python library BeautifulSoup was used to scrape the data from their public online registers.
  • Given the lack of consensus in terms of defining, describing, and grouping types of psychotherapy, in order to compare them and extract key descriptive information I combined articles from multiple mental health-oriented websites . The reason for using multiple descriptive articles for each therapy type was to emphasize key information (that would be repeated in all sources, e.g. main theoretical ideas or techniques used) and reduce the importance of subjective opinions, interpretations, or writer’s bias.

Approach Breakdown

  • Python’s Pandas module was used to extract and aggregate unique types of psychotherapy from BACP and UKCP registers, as well as to count the number of professionals practicing them.
  • Python and Natural Language Processing (NLP) were used to gain insight into key aspects of each therapy based on their descriptions collected from multiple sources. Topic modelling (LDA) and cosine similarity between vectorised descriptions were used to guide the identification of similar approaches and group them together.
  • Venngage.com was used to build infographics for data visualisation. Part 1 outlines the main types and approaches, while Part 2 provides short descriptions of main modalities.
  • Amcharts was used to visualise how often different therapies are used by the same practitioner.
  • Gephi was used to visualise therapies as a network connected by descriptive keywords.

Analysis

1. Extracting therapy types

Figure 1. Reduction and merging of similar therapy typesInitial extraction of unique therapy names from 'Types of Therapies Offered' section of the profiles revealed 130 different values. The list was reduced to 33 by:

  • Merging therapies that appear multiple times under different names (e.g. Jungian therapy, Jungian psychology, analytical psychology, etc).
  • Merging titles which share methodology but differ in training (e.g. integrative counselling, integrative psychotherapy, or integrative psychotherapeutic counselling).
  • Splitting types that are a combination of other therapies (e.g. Systemic and Family therapy would be split into Systemic therapy and Family therapy).
  • Removing types that are complementary therapies used in conjunction to more mainstream therapies (e.g. animal assisted therapy and educational therapy)

Remaining therapy types were separated into two categories:
  • Therapies defined by theoretical approach or methodology (e.g. 'psychoanalysis')
  • Therapies defined by type of client (e.g. 'adolescent therapy')











2. Calculating therapy type availability

Relative 'availability' of therapies was calculated by counting the number of therapists offering each type. Due to the fact that some therapies were mentioned in other parts of the profile (e.g. 'About Me' instead of 'Therapy Types Offered'), entire profiles were scanned for keywords indicating therapy types offered.

Figure 2. Psychotherapy type availability by numebr of practitioners

3. Identifying similarities between therapies

Three main techniques were used to create broader therapy groups:

  • LDA was used to identify hidden themes between the documents.
  • Cosine similarity between vectorised descriptions was used to group ambiguous therapies with those that are most similar.
  • Word frequency and TF-IDF were used to identify key words and ideas for each therapy type. See here for a graph visualising key word network.

Challenges: no single model could be identified with an ideal split of therapy groups sharing an underlying theme. Whereas some groups were easier to idetify (e.g. Unconscious-themed or art/expression-themed (See Figs 3 and 4)), others were more difficult or would only appear with a different set of parameters. Given a significant overlap between the therapy types and terminology/concepts, separation in some cases would prove more difficult. For example, terms such as 'Child' and 'Adult' could mean Ego states in Transcational Analysis, and people in Psychoanalysis. Additionally, therapies such as Psychosynthesis, heavily influcenced by Humanistic and Psychodynamic approaches could potentially be grouped with either of them. In the end, cosine similarity between therapy descriptions was used to guide the group assignments, with types that are more similar to each other being grouped together.

Figure 3. LDA identifying Unconscious-oriented theme. Figure 4. LDA identifying art/expression-oriented theme.

Results

  • 35 main therapy types were identified in UKCP and BACP registered therapists' profiles.
  • 30 of them were best defined by their underlying theory and methodology (e.g. psychoanalysis), while remaining 5 - by client type (e.g. couple therapy or family therapy).
  • LDA revealed that the best grouping of the therapies can be done by using 7-9 main themes.
  • Analysis of the most relevant terms in each group suggested the following underlying themes:
    • Humanism / Empathy / Emotion
    • Thoughts & Behaviour / Structure
    • The Unconscious
    • Relationships / Interactions
    • Spirituality / Philosophy
    • Trauma / Body
    • Art / Expression
  • Given that Integrative Psychotheapy is defined as combination of various methodologies, it was too vague and flexible to be assigned to any group. In fact, being trained in different approaches and use them in therpay for best results is almost the norm now, as vast majority of therapists seem to be working this way (see here for visualisation of this finding).

Putting it all together

1. Main Types

1. Main Modalities and Descriptions