Faculty Profiles

Richard Francoeur

Professor
School of Social Work

Social Work Building 145
516.877.4337
francoeur@adelphi.edu
https://​www.​researchgate.​net/​profile/​Richard_Francoeur

General Information

Personal Statement

Personal Statement

Background

A medical social worker for several years in the VA Pittsburgh Healthcare System, I pursue research and scholarship on middle-aged and older adults; comorbidity and multimorbidity (co-occurring illness conditions, symptoms, or stressors); and chronic, palliative, and supportive care.  I also create or extend innovations in statistical modelling—notably of synergies (curvilinearity and interaction) and subgroups (data-mining, clustering)—as improved approaches to derive knowledge. 

I completed several projects, most on my own, such as 1) symptom clusters of late-life "masked" depression (with low dysphoric—yet also low positive—affect) in progressive vascular conditions (hypertensive, arterial, cardiac, or cerebral), including metabolic syndromes (with excess weight and/or diabetes); 2) sickness malaise in cancer when neuroimmune symptoms interact; and 3) financial strain as age-based reactions by cancer outpatients to objective family financial stressors and their cumulative stress. 

I am leading a project with two Adelphi computer science alumni to build an easy-to-use web application that executes algorithms I developed (published and unpublished) to probe effects of interacting variables. 

I am also developing an improved method of meta-analysis (a quantitative approach to pool and estimate results across a set of studies and to compare study-specific results), which I will use to reanalyze whether breast cancer support groups not only reduce symptoms but lead to improved survival.  This methodological advance will address a pressing need to derive more credible meta-analyses of clinical practices and programs when the evidence is based on heterogeneous participants, study conditions, interventions, and/or research methods.  These more valid findings are important in the process of justifying and adopting sound evidence-based practices, programs, and related health or social policies.

I attracted more than a quarter million dollars in funding (direct costs) as principal investigator of a National Institute of Mental Health grant (R03) on late-life vascular depression, a Hartford Geriatric Social Work Faculty Scholar Award, and an Open Society Social Work Leadership Award (Project on Death in America).

Teaching Interests

Social work practice; social work in chronic, palliative, and end-of-life care

Evidence-based practice; client monitoring with single-case designs

Program development and evaluation; systematic reviews; meta-analysis

Organizational context of social work; social enterprises by social workers

Innovations in Statistical Modelling

I recently forged data-mining of homogeneous subgroups (derived subsets of observations) and their effects from heterogeneous samples, by adjusting multiple regression for heteroscedasticity, multicollinearity, and unspecified predictors (in press). 

To probe comorbidity, co-occurrence, and clustering, I created or extended ways to: 1) detect interactions in moderated regression (MR) more sensitively and precisely (i.e. with lower standard errors and tighter confidence intervals); 2) interpret quickly the nature (magnifying and/or buffering), strength, and statistical significance of MR synergies across discrete values of interacting variables (app pending; this advance avoids the need to re-specify predictors at different levels, re-estimate MR repeatedly, and construct multiple graphs); and 3) unveil elusive, unbiased clusters of psychometric items or symptoms of a latent trait either broadly (across a group) or uniquely (within an MR-targeted subgroup). 

The last of these advances affords insight into distinct presentations, even phenomenologies, tapped by a latent trait that occur across a group (determined by a predictor), or synergistically, within a subgroup (determined by interacting predictors).  It overcomes the potential for common, insidious confounding in regression-based multiple indicators-multiple causes (MIMIC) models, which can only estimate direct (unique) effects of predictors to all but one of the items or symptoms of the latent trait.  Even if it seems justified not to specify the direct effect to a certain item (i.e. it is fixed usually at zero), hidden bias may infect the latent trait and proliferate across items, undermining the validity of specified (shared and direct) effects.

The advance avoids this difficulty by offering a new way to specify a MIMIC model that enables the direct effect on every single item or symptom of a scale or subscale to be unveiled (while still adjusting shared effects between and across the items or symptoms, the latter to account for the level of the latent trait).  It reveals the subset of items or symptoms that have statistically significant direct effects, which comprise the item or symptom cluster within the group or subgroup. 

Links provided to external websites, including personal faculty sites, do not imply an endorsement by Adelphi University of those sites, their content, or associated products and services.