Research report
Hierarchical screening for multiple mental disorders

https://doi.org/10.1016/j.jad.2013.05.085Get rights and content

Abstract

Background

There is a need for brief, accurate screening when assessing multiple mental disorders. Two-stage hierarchical screening, consisting of brief pre-screening followed by a battery of disorder-specific scales for those who meet diagnostic criteria, may increase the efficiency of screening without sacrificing precision. This study tested whether more efficient screening could be gained using two-stage hierarchical screening than by administering multiple separate tests.

Method

Two Australian adult samples (N=1990) with high rates of psychopathology were recruited using Facebook advertising to examine four methods of hierarchical screening for four mental disorders: major depressive disorder, generalised anxiety disorder, panic disorder and social phobia.

Results

Using K6 scores to determine whether full screening was required did not increase screening efficiency. However, pre-screening based on two decision tree approaches or item gating led to considerable reductions in the mean number of items presented per disorder screened, with estimated item reductions of up to 54%. The sensitivity of these hierarchical methods approached 100% relative to the full screening battery.

Limitations

Further testing of the hierarchical screening approach based on clinical criteria and in other samples is warranted.

Conclusions

The results demonstrate that a two-phase hierarchical approach to screening multiple mental disorders leads to considerable increases efficiency gains without reducing accuracy. Screening programs should take advantage of prescreeners based on gating items or decision trees to reduce the burden on respondents.

Introduction

There is a demand for fast and accurate mental health screening in a range of general population settings where time and resource restrictions preclude administration of comprehensive clinical measures. Mental health screening has application to research and to clinical settings, including virtual clinics (Cuijpers et al., 2009, Donker et al., 2009), primary care (Spitzer et al., 1999) and schools (Husky et al., 2011, Weist et al., 2007). Although there are accurate and brief self-report scales available for assessing many specific mental disorders, there is little knowledge about the most efficient methods of screening for multiple disorders (Donker et al., 2009). We propose a new method of screening for multiple mental disorders that combines the strong psychometric characteristics of existing epidemiological screening scales with shorter mean administration time. Specifically, hierarchical screening refers to a multi-phase screening process involving presentation of a very brief screener for general psychological distress, followed by more lengthy screening for respondents who meet specified criteria on the brief screener. This hierarchical selection process has the potential to increase efficiency through the reduction of mean administration time, without loss of precision. Furthermore, hierarchical screening across multiple disorders may generate efficiency gains as a result of the high rates of comorbidity seen among mental disorders (Kessler et al., 2005), which may be overlooked by serial screening methods.

Two-phase screening has been previously used to screen for a single disorder (Clover et al., 2009). The two-phase approach is also cognate to two-phase clinical diagnosis, which is often used in national mental health surveys (e.g., Sunderland et al., 2012), although this approach is not true multi-phase screening as the second phase of assessment is a lengthy diagnostic interview. One previous study developed this joint screener-diagnosis process further, providing a psychometric evaluation of a gating approach to screen for multiple disorders using the DISC Predictive Scales screen (DPS screen; Lucas et al., 2001). This evaluation found greater screening efficiency through the use of gating items followed by a brief screener for each disorder. However, the DPS screen was developed for children and adolescents and requires interviewer administration. In addition, the sensitivity of the screeners was not high for several of the disorders, resulting in missed cases. There has been no further development of this gating approach for the self-screening of community-based populations on the basis of multiple disorders. Moreover, a number of alternative optimisation strategies are yet to be tested, with potential to further decrease total response burden without reducing the sensitivity of the existing measures.

In the current study, four methods of hierarchical screening were tested in two online population samples displaying high rates of psychopathology. Participants in both samples completed all mental health measures. The hierarchical screening methods were therefore developed by estimating the mean number of items that would have been presented to participants, simulated based on the criteria for each hierarchy. The screening methods tested were: (i) no hierarchy (control), (ii) a hierarchy based on K6 scores (Kessler et al., 2002), (iii) a hierarchy based on a decision tree using general psychological distress items, (iv) a hierarchy based on a decision tree using items from disorder-specific scales, and, (v) a hierarchy using gating questions from each of the disorder-specific scales.

The K6 tends to be used to identify those at high risk of disorder. However, there has been little examination of whether the K6 can identify individuals at low risk of disorder, thereby reducing their need to respond to a broader battery of screening measures. To identify a more tailored set of items identifying low risk of disorder, two decision tree approaches were tested. The decision tree is a method to identify subgroups at high risk of a specified, known outcome (see Batterham et al., 2009). This method has not previously been applied to screening, but has the potential to optimally distinguish absence of disorder from presence of disorder by identifying a set of screening items tailored to an individual’s risk for disorder. Similar to the DPS screen approach, a gating item approach was also tested using existing measures that incorporate gating in the identification of disorder. Although the present study focused on four common internalizing disorders, the study aimed to lay the foundation for the development of more comprehensive screeners including additional disorders. It was hypothesized that each of the hierarchical screening methods would lead to a reduced mean number of items presented, with negligible reduction in sensitivity.

Section snippets

Participants and procedure

Australian adults were recruited for an online survey using Facebook advertising during July 2012. The survey was hosted on a secure server at the Australian National University. Surveys were completed in approximately 20–30 min and included online informed consent, the screening measures, a number of other mental health measures, sociodemographic characteristics, and concluded with service referral options. For this initial version of the survey (Sample 1), a total of 1360 surveys were

Results

Characteristics of the two survey samples are shown in Table 1. The majority of the missingness on the demographic variables was attributable to these items being presented at the end of the survey. Sample 1 was recruited using an advertisement that directed respondents directly to the online survey. Sample 2 was recruited with an advertisement that linked to an internal Facebook page that included prominent links to the survey and allowed sharing of the page and link across individuals’ social

Discussion

Typically, screening for multiple disorders has relied on serial presentation of a range of scales that are either brief but imprecise (e.g., Donker et al., 2009) or more accurate but somewhat lengthier (e.g., Kessler et al., 2012, Spitzer et al., 1999). The results of the present study indicate that a hierarchical screening approach can substantially increase screening efficiency without sacrificing precision. The use of gating items was shown to be the most successful method of reducing

Role of funding source

The authors gratefully acknowledge the College of Medicine, Biology and Environment, the Australian National University, which funded this study through an early career fellowship support grant. PB is supported by National Health and Medical Research Council (NHMRC) Early Career Fellowship 1035262. AC is supported by NHMRC Early Career Fellowship 1013199. MS is supported by NHMRC Early Career Fellowship 1052327. HC is supported by NHMRC Fellowship 525411.

Conflict of interest

The authors declare no conflict of interest.

References (23)

  • G.V. Kass

    An exploratory technique for investigating large quantities of categorical data

    Applied Statistics

    (1980)
  • Cited by (8)

    • A computerized version of the Patient Health Questionnaire-4 as an ultra-brief screening tool to detect emotional disorders in primary care

      2018, Journal of Affective Disorders
      Citation Excerpt :

      Our results indicate that this instrument is useful to detect MDD and GAD in PC patients whose GP suspects an ED. However, as several meta-analyses (Manea et al., 2016; Mitchell and Coyne, 2007; Plummer et al., 2016) have suggested, while ultra-short questionnaires are useful as a screening tool to efficiently detect a possible mental disorder, more in-depth tests should be administered, following a hierarchical model (Batterham et al., 2013; Muñoz-Navarro et al., 2017c). In this case, the next logical step would be to administer the complete PHQ-9 or GAD-7 to obtain a more comprehensive clinical picture.

    • Discriminant validity and gender differences in DSM-5 posttraumatic stress disorder symptoms

      2016, Journal of Affective Disorders
      Citation Excerpt :

      To our knowledge, this is the first study internationally to investigate discriminant validity and gender differences in DSM-5 PTSD symptoms in a population sample. This study is part of a larger project involving the development and validation of adaptive mental health screeners (Batterham et al., 2013, 2015). From August through December 2014, a series of paid advertisements were placed on Facebook, via the margin of the account or in the newsfeed, targeting registered users living in Australia, aged 18 years or older.

    • Web-based screening for Panic Disorder: Validity of a single-item instrument

      2015, Journal of Affective Disorders
      Citation Excerpt :

      They are also advantageous for patients: flexible hours, fewer journeys, more confidentiality, and less stigmatization (Donker et al., 2010; Marks and Cavanagh, 2009). PD has been detected through the Internet using lengthy instruments of general health assessment, such as the Patient Health Questionnaire (Batterham et al., 2013), and the Body Sensations Questionnaire (Carlbring et al., 2007), or tools which have been designed for other disorders like the Web Based Depression and Anxiety Test (Farvolden et al., 2003) or the Generalized Anxiety Scale (Donker et al., 2011). Brief and specific screening tools must still be developed.

    View all citing articles on Scopus
    View full text