Cramer et al. model behavioral disorders purportedly using manifest variables (symptoms) rather than latent constructs (diagnostic categories). We challenge the assumption that a symptom is a manifest variable, and use the symptom of sleep disturbance as an example because the target article authors have used this as an exemplar of a symptom shared by several disorders (cf. Diagnostic and Statistical Manual of Mental Disorders (DSM-IV); American Psychiatric Association [APA] 1994). It might seem that a symptom such as sleep disturbance is an empirical observation rather than a theoretical abstraction like depression; but sleep disturbance is itself an abstraction that encompasses multiple, specific instances of sleep loss: delayed sleep onset, mid-sleep and early morning awakening, restless sleep, and disturbed circadian rhythms and sleep stages (APA 1994). To say that someone has a sleep disturbance is to assert that they are disposed to have multiple instances of sleep loss in one or more of these domains.
Conceptualizing symptoms as manifest or latent variables has implications for assessment. The nodes in Cramer et al.'s network were categorically defined symptoms assessed with the Composite International Diagnostic Interview (CIDI) (Kessler & Uslum Reference Kessler and Ustun2004; Kessler et al. Reference Kessler, Chiu, Demler and Walters2005b), a structured clinical interview in which symptoms are assessed in terms of presence/absence rather than with a dimensional score. Clinical interview is one among several methods of assessing sleep disturbance, including questionnaires, diaries, polysomnography (PSG), and laboratory observations, each of which differs in their reliability and validity (Crocker & Algina Reference Crocker and Algina1986). Reliability considers the similarity between observed and true scores. It is necessary to make this distinction if one assumes that the same phenomenon is being measured during multiple observations. In the case of validity, a distinction is made between what is being measured and the instrument that is used to measure it. For example, we might assume that an interview and PSG measure the construct of sleep disturbance, though they might differ in their validity (i.e., the extent to which they are influenced by other factors). Moreover, there is an implicit understanding that the true sleep disturbance score represents a latent variable composed of the observations (manifest variables) of each testing occasion. Hence, the categorical sleep disturbance symptom assessed by the CIDI is assumed to be an abstraction of manifest (observed) instances of sleep loss.
Conceptualizing symptoms as latent constructs has implications for diagnostic validity.
What is the nature of the categorically defined symptom of sleep disturbance as a unit of analysis in diagnostic classification and differential diagnoses of mood and anxiety disorders? The determination that sleep disturbance is a symptom of a mood or anxiety disorder assumes the following: the sleep loss is frequent, substantial, persistent, unintentional; not due to the presence of a medical condition or a substance; and not due to extrinsic (environmental) factors (APA 1994). Insomnia complaints are common in the population, with a one-year prevalence of 30–40% (APA 1994). Hence, ascertaining the frequency, intensity, and persistence of the sleep disturbance is necessary to assess whether it is severe enough to be considered abnormal. Intentional sleep loss (e.g., a student who remains awake all night studying for exams), or sleep loss due to a medical condition, use of substances, disrupted circadian rhythms (e.g., jet lag), or due to environmental factors (loud neighbors), would not be diagnosed as a symptom of a mood or anxiety disorder (APA 1994). We therefore contend that sleep disturbance as a symptom of a mood or anxiety disorder is a latent construct that encompasses manifest observations of unintentional, frequent, substantial, and persistent sleep loss not due to the aforementioned extrinsic factors, and that these manifest variables must be assessed in order to determine whether the sleep disturbance does in fact qualify as a bona fide symptom of a mood or anxiety disorder.
Consider the symptom of sleep disturbance in posttraumatic stress disorder (PTSD). Though this PTSD symptom has many features in common with primary insomnia (e.g., Inman et al. Reference Inman, Silver and Doghramji1990), sleep disturbance in PTSD can be discriminated from that in primary insomnia according to differences in REM-sleep abnormalities (e.g., Inman et al. Reference Inman, Silver and Doghramji1990; Mellman et al. Reference Mellman, Bustamante, Fins, Pigeon and Nolan2002) and frequency of sleep disturbance due to nightmares (Mellman & Pigeon Reference Mellman, Pigeon, Kryger, Roth and Dement2005), due to fear of sleep and of the dark (Inman et al. Reference Inman, Silver and Doghramji1990), or due to nocturnal hypervigilance (e.g., lying awake listening for strange sounds) (Deviva et al. Reference Deviva, Zayfert, Pigeon and Mellman2005). Hence, although the construct of sleep disturbance is found in several behavioral disorders (APA 1994), the observed manifestations of sleep disturbance in PTSD can be discriminated from those associated with other behavioral health conditions. Analysis of sleep disturbance as a latent construct reflecting manifest observations such as, for example, frequency of delayed onset due to fear of nightmares could reduce diagnostic overlap and enhance diagnostic validity. Clinical scientists have also recognized the distinction between sleep disturbances in PTSD and those found in other disorders by developing interventions for PTSD patients that specifically target nightmares and nocturnal hypervigilance (Deviva et al. Reference Deviva, Zayfert, Pigeon and Mellman2005). Hence, recognizing that a symptom reflects a latent construct is not an abstract, semantic issue. Regardless of whether one adopts network or latent construct models of behavioral disorders, a clear understanding of how we define and measure behavioral phenomena, and of the nature of the units of analyses we employ to characterize disorders, has nontrivial implications for determining the best course of treatment, which is the ultimate goal of developing accurate diagnostic methods.
Treating symptoms as proxies for manifest observations leads one to think of them as empirical observations rather than as theoretical constructs. However, assessment, diagnosis, and treatment of behavioral disorders are better served by careful consideration of the basis on which symptoms are identified and measured.
You have
Access
Cramer et al. model behavioral disorders purportedly using manifest variables (symptoms) rather than latent constructs (diagnostic categories). We challenge the assumption that a symptom is a manifest variable, and use the symptom of sleep disturbance as an example because the target article authors have used this as an exemplar of a symptom shared by several disorders (cf. Diagnostic and Statistical Manual of Mental Disorders (DSM-IV); American Psychiatric Association [APA] 1994). It might seem that a symptom such as sleep disturbance is an empirical observation rather than a theoretical abstraction like depression; but sleep disturbance is itself an abstraction that encompasses multiple, specific instances of sleep loss: delayed sleep onset, mid-sleep and early morning awakening, restless sleep, and disturbed circadian rhythms and sleep stages (APA 1994). To say that someone has a sleep disturbance is to assert that they are disposed to have multiple instances of sleep loss in one or more of these domains.
Conceptualizing symptoms as manifest or latent variables has implications for assessment. The nodes in Cramer et al.'s network were categorically defined symptoms assessed with the Composite International Diagnostic Interview (CIDI) (Kessler & Uslum Reference Kessler and Ustun2004; Kessler et al. Reference Kessler, Chiu, Demler and Walters2005b), a structured clinical interview in which symptoms are assessed in terms of presence/absence rather than with a dimensional score. Clinical interview is one among several methods of assessing sleep disturbance, including questionnaires, diaries, polysomnography (PSG), and laboratory observations, each of which differs in their reliability and validity (Crocker & Algina Reference Crocker and Algina1986). Reliability considers the similarity between observed and true scores. It is necessary to make this distinction if one assumes that the same phenomenon is being measured during multiple observations. In the case of validity, a distinction is made between what is being measured and the instrument that is used to measure it. For example, we might assume that an interview and PSG measure the construct of sleep disturbance, though they might differ in their validity (i.e., the extent to which they are influenced by other factors). Moreover, there is an implicit understanding that the true sleep disturbance score represents a latent variable composed of the observations (manifest variables) of each testing occasion. Hence, the categorical sleep disturbance symptom assessed by the CIDI is assumed to be an abstraction of manifest (observed) instances of sleep loss.
Conceptualizing symptoms as latent constructs has implications for diagnostic validity.
What is the nature of the categorically defined symptom of sleep disturbance as a unit of analysis in diagnostic classification and differential diagnoses of mood and anxiety disorders? The determination that sleep disturbance is a symptom of a mood or anxiety disorder assumes the following: the sleep loss is frequent, substantial, persistent, unintentional; not due to the presence of a medical condition or a substance; and not due to extrinsic (environmental) factors (APA 1994). Insomnia complaints are common in the population, with a one-year prevalence of 30–40% (APA 1994). Hence, ascertaining the frequency, intensity, and persistence of the sleep disturbance is necessary to assess whether it is severe enough to be considered abnormal. Intentional sleep loss (e.g., a student who remains awake all night studying for exams), or sleep loss due to a medical condition, use of substances, disrupted circadian rhythms (e.g., jet lag), or due to environmental factors (loud neighbors), would not be diagnosed as a symptom of a mood or anxiety disorder (APA 1994). We therefore contend that sleep disturbance as a symptom of a mood or anxiety disorder is a latent construct that encompasses manifest observations of unintentional, frequent, substantial, and persistent sleep loss not due to the aforementioned extrinsic factors, and that these manifest variables must be assessed in order to determine whether the sleep disturbance does in fact qualify as a bona fide symptom of a mood or anxiety disorder.
Consider the symptom of sleep disturbance in posttraumatic stress disorder (PTSD). Though this PTSD symptom has many features in common with primary insomnia (e.g., Inman et al. Reference Inman, Silver and Doghramji1990), sleep disturbance in PTSD can be discriminated from that in primary insomnia according to differences in REM-sleep abnormalities (e.g., Inman et al. Reference Inman, Silver and Doghramji1990; Mellman et al. Reference Mellman, Bustamante, Fins, Pigeon and Nolan2002) and frequency of sleep disturbance due to nightmares (Mellman & Pigeon Reference Mellman, Pigeon, Kryger, Roth and Dement2005), due to fear of sleep and of the dark (Inman et al. Reference Inman, Silver and Doghramji1990), or due to nocturnal hypervigilance (e.g., lying awake listening for strange sounds) (Deviva et al. Reference Deviva, Zayfert, Pigeon and Mellman2005). Hence, although the construct of sleep disturbance is found in several behavioral disorders (APA 1994), the observed manifestations of sleep disturbance in PTSD can be discriminated from those associated with other behavioral health conditions. Analysis of sleep disturbance as a latent construct reflecting manifest observations such as, for example, frequency of delayed onset due to fear of nightmares could reduce diagnostic overlap and enhance diagnostic validity. Clinical scientists have also recognized the distinction between sleep disturbances in PTSD and those found in other disorders by developing interventions for PTSD patients that specifically target nightmares and nocturnal hypervigilance (Deviva et al. Reference Deviva, Zayfert, Pigeon and Mellman2005). Hence, recognizing that a symptom reflects a latent construct is not an abstract, semantic issue. Regardless of whether one adopts network or latent construct models of behavioral disorders, a clear understanding of how we define and measure behavioral phenomena, and of the nature of the units of analyses we employ to characterize disorders, has nontrivial implications for determining the best course of treatment, which is the ultimate goal of developing accurate diagnostic methods.
Treating symptoms as proxies for manifest observations leads one to think of them as empirical observations rather than as theoretical constructs. However, assessment, diagnosis, and treatment of behavioral disorders are better served by careful consideration of the basis on which symptoms are identified and measured.