Measurement

This page is under construction.

Whether using the IBM-HS, or another theory/model to guide your investigation of the constructs that shape the mental health help-seeking process, it is essential to put considerable thought into your measurement decisions. This page provides a guide to help you think through your goals for a given investigation and what that means for how you should operationalize the help-seeking constructs most relevant to your investigation.

• First, we must define what the exact help-seeking behavior of interest is.
  • Help seeking behavior can be defined, operationalized, and measured in a variety of ways, with each approach accompanied by advantages and disadvantages. Is the behavior attending an initial appointment, which is likely to be an intake appointment that is often different in format from subsequent working treatment sessions? Is the behavior attending multiple appointments or perhaps completing the entire designated course of treatment? Is the behavior simply taking concrete steps toward attending an appointment?  Is it “schedule an initial appointment” via electronic (online form, email, text) or phone (call) with a provider? How do we handle the fact that most people do not directly reach out to a mental health provider/agency but instead their first professional point of contact is their primary care physician, who may or may not provide a referral in one form or another to a behavioral healthcare provider/agency? Is the behavior an even earlier step such as talking with a trusted confidant or consulting a resource (e.g., brochure, website, app) to gather information about potential treatment options or treatment providers? Or about a specific identified provider?
  • The help-seeking process is an iterative, messy care pathways journey and there are a variety of steps one must take to showing up for an appointment, and the journey is often not linear and oftentimes progress is met with regress.
• Second, we must define the target/source of the help-seeking behavior.
  • Will items be framed in terms of seeking help from one particular kind of professional (e.g., psychologists), from any one of several professionals (e.g., psychologists, counselors, licensed clinical social workers), from “mental health professionals” generically?
    • How are these professionals named, described, and defined, or not, for respondents? Real people often do not know, and may not particularly care about, the similarities and differences among these professionals. What does this mean for your investigation? People may have access to some of these types of professionals and not others, depending on their geographic location and socioeconomic means. What will this mean?
    • What boundaries will be set for which professionals get counted as behavioral healthcare professionals? Do pastoral counselors get included? Do traditional indigenous healers get counted? Do professionals who get some auxiliary training in mental health get included? Do non-psychiatrist physicians get included?
• Third, we must define the type of help the behavioral healthcare professional is thought to (or would actually) provide. Are we interested only in seeking talk therapy? Medication? Alternative or holistic treatments? Whether we leave this unspecified, or specify it in some manner, there are advantages and disadvantages to all choices.
• Fourth, we must define the reason/motivator for seeking professional behavioral healthcare.
  • Is the person seeking help for what mental health professionals would consider is a recognized DSM-5 disorder? What about sub-clinical presenting concerns such as general distress? What about mild vs. moderate vs. severe distress and how that relates to outpatient versus inpatient versus coerced/mandated/involuntary commitment aspects of more severe distress?
  • Who defines what the reason/motivator is?  Professionals with their formal diagnostic taxonomies? Respondents who may use lay or culture-bound conceptualizations of their distress? This changes how the issue is described.
  • People’s help seeking behavior, intention, attitudes, etc. may be different depending on what the reason in question is.
  • It is possible to frame items generally in terms of “for a mental health concern” and then define that to include a variety of issues?  It is possible to describe a set of symptoms or functional impacts? It is possible to state the name of a given disorder and possibly also define that for respondents?
• Fourth, we must define the population of interest.
  • Different groups of people in different circumstances will have different relationships to reasons/motivators, types of mental health professionals, etc. The things that influence help seeking will vary across populations.  There may be some things that we tend to see as mattering across a variety of populations, but there are also some things that will matter for some populations but not others. Thus, the wider the group you are trying to generalize your findings to, the more those findings will be most applicable to those who numerically dominate the sample and the less they will be applicable to those in the numerical minority. We can think about populations in terms of contexts, locations, cultural identity background, etc.  You will have to set the population boundary conditions at some level.
  • We can also think about the current mental health status (e.g., distressed, depressed) as defined by _________ (lay self-perception, formal diagnosis that may have been offered by prior professional they had contact with, a validated screening measure) and current treatment status as further defining the boundaries of the population of interest.  We’ll say more about this in the next section.
• Fifth, we must think about whether we want to use a standard frame or a hypothetical frame for our study.
  • A given study may use an overall frame and also measure-specific frames that are consistent or vary across parts of the study. A standard frame (that uses “unconditional language”) asks the respondents to describe their thoughts, feelings, intentions, and behaviors related to seeking professional behavioral healthcare given their current mental health status (non-distressed, distressed, etc.). Questions about attitudes, intention, etc. can be framed in terms of what they truly believe/experience/intend right now given their current mental health status.
    • This requires that questions be relevant to all respondents’ current mental health status.  For example, asking someone how much stigma they feel about their mental illness requires them to identify as having a mental illness. As another example, asking someone how they would feel about their seeking help from a mental health professional presupposes that seeking help would be a relevant behavior given their circumstance. Generally, when seeking to understand the relationship between two variables, it is best when both variables have a normal distribution of responses on them from the population, which means most of the sample should be answering toward the middle of that scale. Therefore, when it comes to a variable like intention to seek help, it makes sense that all respondents presented with the question of “I intend to seek help” on a “strongly disagree to strongly agree” scale could potentially have a reason to indicate intention to seek help.  For one to intend to do a behavior, the behavior must be relevant to them.  By way of analogy, if a researchers wanted to study the predictors of seeking grief counseling, it would make sense to restrict the sample to people for whom grief counseling is relevant (e.g., people who have experienced a significant loss).In summary, when using standard frame, it is important that samples are restricted to people who are experiencing a mental health concern of some kind. As noted above, one should be intentional about how “have a mental health concern” is determined—whether through self-report scores on a screening measure, through self-identification as having a mental health concern or being generally distressed, or through the clinical decision of a trained mental health professional doing a diagnostic interview. Thus, use of standard frame items requires that study advertising, participant recruitment, and participant screening are effective at only ushering in qualified participants to complete the help seeking items.
    A hypothetical frame (that uses “conditional language”) asks respondents to answer based on how they imagine they would think, feel, and behave in a hypothetical situation. 
    • The core of the hypothetical situation typically focuses on the respondents’ mental health status, asking them to imagine that they are at present experiencing a mental health concern.  This concern can be defined or illustrated (e.g., description of symptoms, functional impact, duration) at some chosen degree of detail. The aim is to help participants richly imagine the situation you are wanting them to imagine, as the vividness and reality of that hypothetical picture will help them to make accurately make guesses about how they would think, feel, and behave in this hypothetical scenario. The harder it is for them to imagine the hypothetical situation, the arguably less accurate their anticipated thoughts, feelings, and behaviors are likely to be to what would actually happen in real life. The advantage of hypothetical frame is that, assuming the hypothetical framed items can be answered in a manner that reflects how they would truly respond if the hypothetical situation was genuinely true for them at the time of the survey, we can administer these items to persons who are not currently distressed, including persons who might become distressed in the future but are not currently and for whom we would like to get a prediction of how they may think, feel, and behave if they find themselves in the future to be dealing with a mental health concern. It allows us to sample from a winder range of the cultural population of interest, not just those within that sample who are currently distressed.
  • Some validated instruments are framed in a standard frame, some in a hypothetical frame, and some instruments can use either frame by either changing the item wording, changing the instrument instructions wording, or changing the survey introduction wording.
• Sixth, we must carefully select help-seeking outcome variables to be measured.
  • The gold standard outcome measure in help seeking research is help seeking behavior. However, it can be challenging to measure help-seeking behavior in terms of cost to researchers (e.g., money, resources, time) and participants (e.g., showing up physically to a research lab for some experimental studies, responding to multiple surveys for longitudinal studies). Therefore, researchers have also been interested in using alternative outcome variables that are suitable proxies for help-seeking behavior. Research on understanding the precursors to human behavior indicates that people may take action out of habit, due to being triggered by situational cues, and/or because they form a conscious intention to perform a behavior and then may follow through on that behavior. The IBM-HS asserts that one’s intention to seek help is the most proximal precursor to future help seeking behavior, assuming the behavior is under volitional control, the person has relevant knowledge and skills necessary to perform the behavior, and there is an absence of serious environmental constraints that would prevent the performance of the behavior, among other caveats. Indeed, intention is one of the best predictors of prospective behavior within (Adams et al., 2022; Hammer et al., 2018) and outside (Armitage & Conner, 2001; Fishbein & Ajzen, 2010) of the mental health help-seeking context (Hammer et al., under review, p. X).
• • Thus, some researchers prefer to use help seeking intention as the outcome variable of interest, as it is a reasonable proxy for future behavior and has the advantage of allowing for collecting data in a single cross-sectional survey. Therefore, researchers should think carefully about whether they want to measure prospective behavior, intention, or even one of the help-seeking mechanisms (e.g., help-seeking attitude) as the primary outcome variable of interest.
• Seventh, we must decide how to measure help-seeking behavior.
  • Once help-seeking behavior has been defined per the first, second, and third steps outlined above, one must decide how to measure behavior.
  • Behavior is often measured with regarding to four temporal reference points: past help seeking behavior, present help seeking behavior, imminent help seeking behavior, and prospective help seeking behavior.
  • Past help seeking behavior: some researchers ask people if they have sought help in the past, whether that be at any point in their lifetime, the last 6 months, or some other past time frame. Past experience with a behavior influences present thoughts, feelings, and behaviors related to that behavior. Thus, a person who reports having sought help in the past can be understood to have current attitudes, personal agency, and intentions that reflect the lessons learned from that past experience. Because past behavior can change present cognitions, it is logically invalid to treat past help seeking behavior as a dependent variable that is being predicted by current perceptions.  It is OK to use independent variables that would (typically) remain unchanged as a result of past experience with help seeking (e.g., age, gender, personality) as those things don’t violate the causal logic, but it is a bad idea to use social-cognitive variables (stigma, conformity to masculine norms, help-seeking self-efficacy) that are not immutable and trait-like as predictors, as it violates temporal assumptions about causality. Unfortunately, there are several published studies that have made this mistake and it renders those results suspect because the causal directionality of that relationship is ambiguous.
  • Present help seeking behavior: some researchers may ask if a respondent is currently seeking help from a mental health professional, meaning that they are currently experiencing a course of treatment with a professional. Just as cross-sectional designs don’t allow firm conclusions about causality, present help seeking behavior’s relationship with non-trait social cognitive variables violate assumptions regarding causality.
  • Imminent help seeking behavior: some researchers may wish to provide respondents with the opportunity to do (or not) a help seeking behavior as a part of the study protocol. For example, they may offer the respondent an opportunity to see their scores on a mental health screening measure, click on a link to an online mental health provider where they can sign up for an account, click on a link to schedule an appointment with a campus mental health provider, or if in the lab they might be offered an opportunity to have a conversation with a trained mental health practitioner in the next room.  If participants filled out the survey items in advance of being presented with this behavioral choice point, then the causal sequence integrity is maintained and we can treat their prior responses as causal predictors of their imminent help seeking behavior being measured in this study.
  • Prospective help seeking behavior: some researchers may follow up with time 1 participants at a later point in time (e.g., time 2 follow-up survey) to ask them to provide data regarding their subsequent help seeking behavior (or lack thereof) in a self-report fashion.  Other researchers may obtain information on prospective help seeking behavior through a third-party who the respondent has given consent to (e.g., campus counseling center records). But usually time 2 data collection consists of asking the respondent if they have sought behavioral healthcare since the date they completed the time 1 survey. It is possible for respondents to answer incorrectly if their believe they have (not) sought help in the manner defined by the researchers despite meeting researchers criteria. More likely, it is possible for respondents to answer in a socially desirable manner, possibly indicating they have sought help when they have not or vice versa.  However, most researchers tend to treat their self-report responses as accurate, for practical reasons. In this prospective design
  • It is common for researchers to always measure past help-seeking behavior regardless of whether they will also measure behavior in another way, but use it to characterize the sample and potentially use it as a predictor of present thoughts, feelings, and intentions regarding help seeking, rather than use it as a DV.
  • I recommend always measuring past help-seeking behavior, as it informs help-seeking precursors, help-seeking mechanisms, and outcomes alike.
• Eight, having selected the outcome variables of interest, we must also determine what help-seeking precursors and help-seeking mechanisms we also wish to include to help answer the particular research questions we have.
  • What these variables are is dictated by your area of focus and often what theory of behavior/help-seeking is grounding your work.
  • For example, IBM states that (a) help-seeking precursors include constructs such as structural forces, cultural influences, individual differences, mental health status and perceptions, past experiences with mental health help seeking, environmental constraints, and knowledge and skills and (b) the help-seeking mechanisms are attitude, perceived norm, and personal agency.
• Ninth, you must select appropriate measures for all constructs of interest.
  • The theory or theories you are using may provide guidance on how the constructs specified by that theory should be measured. Grounding your decisions in a clear theory or model is important because different models/theories can have differing conceptualizations of the same construct, which can lead to different measurement approaches, which can lead to confusion and inaccuracy when trying to compare results across studies and draw conclusions. This is where concepts such as construct proliferation (see Spiker & Hammer, 2019) become important to consider.
  • We want to avoid measuring constructs in a manner that deviates from the chosen theory/model, avoid using measures with problematic psychometric properties, and avoid using measures that are not congruent with the decisions we made regarding our definition of help-seeking, the type of help, reason, framing, population, etc.  For example, as noted by Hammer and colleagues (under review): “one of the most popular help-seeking attitude measures, the Attitudes Toward Seeking Professional Psychological Help – Short Form (ATSPPH-SF; Fischer & Farina, 1995), was historically used by researchers seeking to ground their work in the reasoned action tradition. However, rather than only measuring construct-relevant variance in attitude (as defined by the reasoned action tradition), this instrument’s mean score also measures construct-irrelevant variance in knowledge and intention (Hammer et al., 2018)” (p. X). Thus, the ATSPPH-SF is not an appropriate measure when seeking to measure attitude in the precise manner defined by the IBM-HS and the wider reasoned action tradition.
• In conclusion, having done these 9 steps outlined above on this webpage, we are well positioned to conduct a high quality research study that can identify what mutable help-seeking precursors influence prospective help seeking behavior and/or its proxies, and what mediational mechanisms may account for the influence of these mutable precursors, for a particular population in a particular professional treatment context from a particular provider type.