A. My Research, Scholarship, and Innovations
I study aging, comorbidity, and palliative care (e.g., vascular depression, pain and symptom clusters, financial burden). I devised methods that 1) detect interactions better or interpret them easier in multiple regression (app in development) and 2) link multiple regression fully and without bias to confirmatory factor analysis by making it possible to estimate all causal paths to a latent construct and its observed items. These advances reveal co-occurring variables that are synergistic or form unique psychometric profiles in subgroups.
My research, scholarship, and innovations increase our understanding of older, middle-aged, and underserved adults with chronic medical conditions:
— I detect and interpret symptom clusters (multiple symptoms of an illness in the same person) for insights into palliative care and related mental health care
— I detect and explain comorbidities (multiple illnesses in the same person) and other co-occurrences (for example, the objective stress and perceived strain of financial burden) that reveal patient subgroups at risk of inadequate healthcare
— I improve analytic methods to study synergy (statistical interactions) and unique psychometric profiles in these situations
— In other work, I develop original ideas that create or expand social work roles to improve illness-informed practice, transdisciplinary care, and disease management in community and healthcare settings
My research reveals components of symptom clusters, comorbidities, or other co-occurrences that interact with each other. These interactions reveal synergistic effects that magnify or buffer relationships to health or mental health outcomes, relative to the broader context where these predictors are scattered across the sample without necessarily co-occurring in the same person. It is practical to target synergistic effects since multidisciplinary healthcare teams will have more incentive to screen for them — and not merely co-occurring symptoms or illnesses — as their deleterious or protective effects become realized and as evidence distinguishes circumstances that call for medical versus mental health treatment or both.
My research emphasizes synergistic effects in cancer, diabetes and excess weight, and progressive cerebrovascular disease. These recent or current projects apply innovations I developed to statistical methods or models for deriving more valid findings. They aim to:
— improve the detection of interactions in multiple regression among physical symptoms that constitute symptom clusters, as well as the interpretation of their synergistic effects on comorbid mental health outcomes
— estimate exhaustively specified latent trait models in which progressive cerebrovascular conditions interact with comorbidities or co-occurrences to predict psychometric profiles (symptom clusters) of late-life depression within these synergistic subgroups
These new developments have potential for broad application in health, aging, social work, and other fields.
My research and scholarship fall within three interrelated areas (for each area, I list key achievements, and in the third area, also current projects):
1. Older and middle-aged adults coping with medical conditions or related physical symptoms who present with depression
— I focus on co-occurring illness conditions, symptoms, or psychosocial factors that interact to magnify their relationships to depression, or to reveal distinct profiles of depressive symptoms. My studies reveal that depression may be recognized or hidden (i.e., masked); screened using a single flexible item; experienced as sickness malaise when symptoms of cancer occur in pairs or clusters; and characterized by distinct profiles of depressive symptoms in subgroups of older adults with progressive cerebrovascular disease, diabetes, and/or excess weight (see "Research Interests," sections 1a, 2a, and 2b).
Andrea M. Barsevick, a prominent investigator of cancer symptom clusters, claims my initial study of symptom clusters (Francoeur, 2005; see "Refereed Articles") "...provides the first evidence that symptom pairs can have a synergistic or interaction effect in predicting patient outcomes" [based on the Symptom Management Model and the Theory of Unpleasant Symptoms]. Further, she maintains that the study presents "...the first test of the sickness behavior hypothesis" [in which an immune response, signaled by fever, triggers and sustains physical symptoms and depression known as sickness malaise. These quotations appear on page 975 in: Barsevick, A. M. (2007). The elusive concept of the symptom cluster. Oncology Nursing Forum, 34(5), 971-980]. Recently, I reported several symptom clusters that consistently support the role of fever in the sickness behavior hypothesis, which are distinguished from symptom clusters of the same symptoms in the absence of fever (i.e., the depression hypothesis; Francoeur, 2015 "Part II").
2. Hidden or emerging clinical issues in older and underserved populations with chronic illness, especially during palliative care
— A major focus in this area has been the use of indirect indicators to reveal whether subgroups receiving palliative care may be at risk of forgoing essential health care and necessities. For instance, I published several articles regarding the moderating influence of age on objective financial stress-subjective financial strain relationships. I revealed that the indicator of financial strain from difficulty paying bills, when used alone, would be misleading to identify older outpatients receiving palliative radiation who may be at risk. Compared to younger outpatients incurring similar levels of financial stress, older outpatients were found to report less difficulty paying bills, even as they were more concerned about a second indicator of financial strain, the adequacy of their financial resources and insurance to meet future health needs. This response pattern suggests older outpatients may be more likely to forgo care and necessities that would pose financial hardship. I qualified these relationships further based on the experience of a recent work transition or the number of days impaired by disability (see "Research Interests," sections 3c and 3d).
Through this work, I pioneered conceptual and methodological developments, and deepened empirical understanding, in testing age-related relationships between an overall index of objective family financial stress and indicators of subjective patient financial strain, based on a series of articles regarding financial burden in outpatients initiating palliative radiation (Francoeur, 2007, 2005, 2002, 2001). This foundational work has been an important influence in the conceptualization, design, and analysis of financial stress and strain in later studies of cancer survivors and other populations. A recent study by Sharp and Timmons (2015), for instance, cited my 2005 study eight times to guide their development of financial variables and inform their methods and interpretations.
— I published agency, hospital, and community programming strategies that 1) engage agency social workers to monitor hypertension (Francoeur, 2010); 2) expand the palliative social work role in spirituality to facilitate coping with chronic illness and self-determination in advance care planning (Francoeur, Burke, and Wilson, in press), and 3) integrate police, pharmacists, medical providers, and social workers to ensure safe access to medication for palliative care while preventing prescription drug abuse (Francoeur, 2011). In the latter area, my commentary has been downloaded over 1,400 times. Among the sources that cite it are the American Society of Clinical Oncology website ("Cancer pain management: Safe and effective use of opioids," 2015 Educational Book, http://meetinglibrary.asco.org/content/11500593-156) and the Oxford Textbook of Palliative Nursing ("Chapter 7: Pain at the end of life," https://books.google.com/books?isbn=0199332347).
3. Creating innovations to statistical methods and models I use in my work that may influence research in health, aging, social work, and other areas
I created three innovations to statistical methods or models. I published two of the innovations to improve detection and interpretation of statistical interaction effects in multiple regression, a common statistical approach in health and aging research and many other disciplines. Both innovations stem from my work on cancer symptom clusters. They can be used together to discover and interpret influential pairs and clusters with synergistic — and not merely co-occurring — components based on symptoms, psychosocial factors, or disease markers. Currently, these two interrelated innovations are being programmed into an app to facilitate wider access by researchers (see "Research Interests," section 1b).
I am writing articles that will demonstrate a third innovation from my research on at-risk subgroups of older adults with progressive cerebrovascular conditions. These cerebrovascular conditions interact either with other comorbid conditions, such as diabetes and/or excess weight, or co-occurring psychosocial or developmental factors, such as education level, and these interactions predict unique profiles, or symptom clusters, of depressive symptoms from the CES-D Depression Inventory. This third innovation constitutes a unique modeling specification I discovered to estimate exhaustively specified latent trait models that combine confirmatory factor analysis with multiple regression. In addition, I am writing a grant proposal to demonstrate a fourth statistical innovation.
I now describe each of these statistical innovations:
— Sequential Residual Centering (SRC) is one of two statistical innovations I derived from my methodological work on cancer symptom clusters. SRC is a breakthrough because it overcomes low sensitivity in multiple regression to detect terms that involve interactions among predictor variables. This dilemma has challenged researchers ever since computer software for multiple regression became available in the 1960s. I also derived extended versions of SRC for further improvements in the sensitivity of moderated regressions with control, secondary, or curvilinear (quadratic) predictors (Francoeur, 2013, 2015 Part I, 2015 Part II; my open-access statistical article and both parts of my symptom cluster application were downloaded over 1,300 times each).
In addition, I devised a strategy based on the SRC to calculate the Total Net Moderator Effect, which reveals whether the buffering or magnifier effect is stronger in the sample, by aggregating each type of effect and comparing them. This strategy may be the only way to interpret complex interactions with more than three linear components or multiple curvilinear components. It also offers insight into the changing strength or direction of the Total Net Moderator Effect when different components of the interaction are considered to be the primary (non-moderating) variable. Finally, when the EZSC post-hoc procedure (in the next bullet) reveals both buffering and magnifier effects from an interaction, the strategy may be used to deduce which effect is stronger, since the stronger effect has the same nature as the Total Net Moderator Effect (Francoeur, 2013).
— The Extended Zero Slopes Comparison (EZSC) is the other statistical innovation I derived from my methodological work on cancer symptom clusters. EZSC extends an original algorithm, the Zero Slopes Comparison (ZSC), which is a post-hoc procedure to interpret interactions between two linear variables detected in multiple regression. EZSC offers new options to reveal the nature of moderator effects if one of the two interacting variables is not linear but quadratic (squared) in its influence, resulting in a curvilinear interaction, or if three linear variables interact. Both ZSC and EZSC overcome the need to re-estimate multiple regressions and construct related graphs to reveal the nature of moderator effects (magnifier and/or buffering). These graphs are not exhaustive; they are limited in displaying "snapshots" of moderator effects at pre-selected values of the predictor variables comprising the interaction. In contrast, both algorithms yield fewer interpretations of aggregate moderator effects across predictor ranges, and not at pre-selected values (Francoeur, 2011a).
At present, I am working with a computer scientist to develop a web app to calculate the reduced standard errors and resulting larger t-values of SRC regression coefficients, along with easy application of the ZSC and EZSC algorithms to interpret the nature and relative strength of statistical interactions detected in raw or SRC multiple regression. This web app will have potential for broad use across areas of research in gerontology and other fields whenever there is a need to interpret synergistic interactions among variables.
— I recently developed a third innovation that constitutes a modeling specification strategy to overcome a critical limitation in deriving valid findings from exhaustively specified models (i.e., all possible relationships are estimated) combining confirmatory factor analysis with multiple regression. These structural equations models are appealing to identify unique psychometric presentations or profiles within subgroups of participants because they simultaneously estimate the factor analysis of a latent construct (e.g., total depression) on its observed "multiple indicators" (i.e., the set of measured psychometric items) while regressing the latent construct and its multiple indicators on a set of "multiple causes" (i.e., predictor variables). However, when exhaustively specified, these multiple indicators-multiple causes (MIMIC) models lack sufficient exogenous information to yield valid estimates and are therefore under-identified. Prior to my new strategy, this limitation created a difficult dilemma for my work since only just-identified ("saturated") or over-identified models can ensure valid detection of unique psychometric profiles within subgroups of participants. I am writing articles about at-risk subgroups of older adults with unique profiles, or symptom clusters, of depressive symptoms when progressive cerebrovascular conditions interact with other comorbid conditions, such as diabetes and/or excess weight, or with co-occurring psychosocial or developmental factors, such as education level. Some profiles suggest distinct symptom phenomenologies of depression in the context of cerebrovascular conditions (see "Research Interests," section 2a).
— I am writing a grant proposal to develop an approach I conceived for distinguishing causality from reverse-causality in multiple regression to target specific contexts that call for medical versus mental health treatment, or both, to relieve symptom feedback loops that link pain to comorbid depression or anxiety in heart failure, lung disease, or multiple organ system failure.
Finally, numerous other scholars have cited my articles and book chapters (Google and Google Scholar reveal 261 works, of which 75 cite one of my symptom clusters studies 105 times), and in two discussion forums for specific statistical software (Stata, SPSS). Although not an exhaustive record, Google Books shows actual citations to my work that appear in 8 books in social work or social welfare, 19 books in health, medicine, or nursing, and 3 books in social science. I am delighted that my findings and innovations are influencing investigators in health, aging, social work, and other disciplines.
I maintain an active presence on ResearchGate, a social networking site for researchers and scientists (https://www.researchgate.net/profile/Richard_Francoeur). I have responded to several questions on ResearchGate from international researchers concerning multiple regression, including detecting and interpreting interactions. For a more extensive discussion of my research, scholarship, and innovations, see "Research Interests" below.
B. My Clinical and Teaching Experiences
My experiences in social work practice and evaluation continue to influence the development of my substantive and methodological interests in health and aging, as well as enrich my teaching and field advising. Foremost among my clinical positions, I served a diverse group of veterans and families as a medical social worker at the VA Pittsburgh Healthcare System. For more than seven years, I gained invaluable clinical experience across a wide range of healthcare settings — ambulatory, outpatient, inpatient, intensive care, physical rehabilitation, dementia care, and community nursing homes.
During this odyssey, I met with patients and family caregivers who were grappling with serious medical conditions, physical symptoms, and care demands. At times, hidden mental health conditions weakened coping and the effectiveness of care. For instance, some older adults with good cognitive functioning seemed withdrawn and disinterested, but denied sickness malaise or feeling blue. This response pattern deterred relatives and health providers from detecting "masked" depression, which may be a sign of illness exacerbation and risk for inpatient readmission, as well as an indicator of compromised safety, care adherence, and spousal caregiving at home. I also became intrigued that in response to screening items, patients and spouses reported lower financial strain when they were older, despite experiencing similar levels of economic stress as patients and spouses at younger ages. These limitations in detecting depression or financial difficulties in older adults sparked my interest in the phenomenon of biopsychosocial issues that remain hidden despite assessment.
I remain committed to the process of revealing hidden or emerging needs, issues, and strategies. Master's and doctoral students in my classes, and Master's students and agency supervisors I advise in field placements, inform me about needs and issues in client populations and the evolution of service delivery at community agencies. I share insights with them from my own clinical, programming, and research experiences, and from my published scholarship, when there may be implications for social work roles, program development, or evaluation. In the classroom, I encourage students to engage each other and me in mutual learning. This shared process is similar to how social workers interact with clients and co-workers in clinical situations, and with colleagues at meetings, trainings, and conferences. I acknowledge when fresh perspectives emerge from mutual learning and try to incorporate some of them into my activities.
In the Master of Social Work (MSW) program, I am currently teaching the first-year course, Foundations of Social Work Practice I, and the second-year advanced elective, Practice in End-of-Life Care. Recently, I revised the latter course to incorporate a greater focus on palliative care and advance care planning as part of the illness trajectory that may lead eventually to the end-of-life, as well as distinct interventions such as existential psychotherapy and forgiveness therapy. In previous years, I taught the first-year course, Human Behavior Theory for Social Work Practice I, and the second-year seminar, Contemporary Social Work. In the latter course, advanced Master's students select and research a specific social problem, purposely integrating knowledge across the four major areas of the curriculum (social work practice; human behavior theory, assessment, and diagnosis; social policy and organizations; and social work research).
In the social work doctoral (Ph.D.) program, I teach a course in Program Development and Evaluation. Over time, I have supplemented the traditional focus on crafting and critiquing program logic models with new course content and assignments to prepare doctoral students to engage in evidence-based practice, including evidence-based program development and evaluation. Doctoral students learn methods for critiquing and conducting systematic reviews and meta-analyses of the empirical literature, and they are introduced to mixed-methods evaluation (i.e., integration of qualitative and quantitative methods within the same study). In previous years, I occasionally taught a course in Advanced Research Topics.