Biological aging is revealed by physical measures, e.g., DNA probes or brain scans. In contrast, individual differences in mental function are explained by psychological constructs, e.g., intelligence or neuroticism. These constructs are typically assessed by tailored neuropsychological tests that build on expert judgement and require careful interpretation. Could machine learning on large samples from the general population be used to build proxy measures of these constructs that do not require human intervention?
Here, we built proxy measures by applying machine learning on multimodal MR images and rich sociodemographic information from the largest biomedical cohort to date: the UK Biobank. Objective model comparisons revealed that all proxies captured the target constructs and were as useful, and sometimes more useful, than the original measures for characterizing real-world health behavior (sleep, exercise, tobacco, alcohol consumption). We observed this complementarity of proxy measures and original measures at capturing multiple health-related constructs when modeling from, both, brain signals and sociodemographic data.
Population modeling with machine learning can derive measures of mental health from heterogeneous inputs including brain signals and questionnaire data. This may complement or even substitute for psychometric assessments in clinical populations.

OBJECTIVE: Proxy reports are often used when patients are unable to self-report. It is unclear how proxy measures are currently in use in adult health care and research settings. We aimed to describe how proxy reports are used in these settings, including the use of measures developed specifically for proxy reporting in adult health populations.
METHODS: We systematically searched Medline, PsycINFO, PsycTESTS, CINAHL and EMBASE from database inception to February 2018. Search terms included a combination of terms for quality of life and health outcomes, proxy-reporters, and health condition terms. The data extracted included clinical context, the name of the proxy measure(s) used and other descriptive data. We determined whether the measures were developed specifically for proxy use or were existing measures adapted for proxy use.
RESULTS: The database search identified 17,677 possible articles, from which 14,098 abstracts were reviewed. Of these, 11,763 were excluded and 2335 articles were reviewed in full, with 880 included for data extraction. The most common clinical settings were dementia (30%), geriatrics (15%) and cancer (13%). A majority of articles (51%) were paired studies with proxy and patient responses for the same person on the same measure. Most paired studies (77%) were concordance studies comparing patient and proxy responses on these measures.
CONCLUSIONS: Most published research using proxies has focused on proxy-patient concordance. Relatively few measures used in research with proxies were specifically developed for proxy use. Future work is needed to examine the performance of measures specifically developed for proxies.
UNASSIGNED: PROSPERO No. CRD42018103179.

In light of the many robust quantitative data sets that include information on attitudes and behaviors related to intimate partner violence (IPV), and in an effort to expand the evidence base around social norms and IPV, many researchers construct proxy measures of norms within and across groups embedded in the data. While this strategy has become increasingly popular, there is no standardized approach for assessing and constructing these norm proxies, and no review of these approaches has been undertaken to date. This study presents the results of a systematic review of methods used to construct quantitative proxy measures for social norms related to IPV. PubMed, Embase, Popline, and Scopus, and PsycINFO were searched using Boolean search techniques. Inclusion criteria comprised studies published since 2000 in English that either (i) examined a norm proxy related to gender or IPV or (ii) analyzed the relationship between a norm proxy and perpetration of, experiences of, or attitudes toward IPV. Studies that employed qualitative methods or that elicited direct measures of descriptive or injunctive norms were not included. Twenty-six studies were eligible for review. Evidence from this review highlights inconsistencies in how proxies are constructed, how they are assessed to ensure valid representation of norms, and how researchers acknowledge their respective method\'s limitations. Key processes and reflections employed by some of the studies are identified and recommended for future research inquiries.

to compare care staff proxies with care home residents\' self-assessment of their health-related quality of life (HRQoL).
we assessed the degree of inter-rater reliability between residents and care staff proxies for the EQ-5D-5L index, domains and EQ Visual Analogue Scale at baseline, 3 months and 6 months, collected as part of the PATCH trial. We calculated kappa scores. Interpreted as <0 no agreement, 0-0.2 slight, 0.21-0.60 fair to moderate and >0.6 substantial to almost perfect agreement. Qualitative interviews with care staff and researchers explored the challenges of completing these questions.
over 50% of the HRQoL data from residents was missing at baseline compared with a 100% completion rate by care staff proxies. A fair-to-moderate level of agreement was found for the EQ-5D-5L index. A higher level of agreement was achieved for the EQ-5D-5L domains of mobility and pain. Resident \'non-completers\' were more likely to: be older, have stayed a longer duration in the care home, have lower Barthel Index and Physical Activity and Mobility in Residential Care (PAM-RC) scores, a greater number of co-morbidities and have joined the trial through consultee agreement. Interviews with staff and researchers indicated that it was easier to rate residents\' mobility levels than other domains, but in general it was difficult to obtain data from residents or to make an accurate proxy judgement for those with dementia.
whilst assessing HRQoL by care staff proxy completion provides a more complete dataset, uncertainty remains as to how representative these values are for different groups of residents within care homes.

In the absence of direct measurements of state-level household gun ownership (GO), the quality and accuracy of proxy measures for this variable are essential for firearm-related research and policy development. In this work, we develop two highly accurate proxy measures of GO using traditional regression analysis and deep learning, the former accounting for non-linearities in the covariates (portion of suicides committed with a firearm [FS/S] and hunting license rates) and their statistical interactions. We subject the proxies to extensive model diagnostics and validation. Both our regression-based and deep-learning proxy measures provide highly accurate models of GO with training R2 of 96% and 98%, respectively, along with other desirable qualities-stark improvements over the prevalent FS/S proxy (R2 = 0.68). Model diagnostics reveal this widely used FS/S proxy is highly biased and inadequate; we recommend that it no longer be used to represent state-level household gun ownership in firearm-related studies.

OBJECTIVE: To provide evidence regarding the construct validity of a sexual aggression proxy in which male participants go on multiple virtual dates with a woman. A unique strength of this proxy is participants\' opportunity to make choices throughout the simulation about how they interact with their virtual date. These decisions determine their exposure to the female agent\'s sexual refusals.
METHODS: Piloting included focus groups (n = 82), surveys (n = 95), and cognitive interviews (n = 32). To establish construct validity, 87 male participants completed two separate sessions: 1) an online survey with measures of discriminant and convergent validity; and 2) a lab session in which they went on multiple dates that included nonsexual options, sexual options which the female agent accepted, and sexual options she refused. Sexual aggression was operationalized as the total number of sexual refusals that participants\' received.
RESULTS: There was strong correlational evidence for discriminant and convergent validity. As hypothesized, there were some differences in the risk factors associated with refusals received on casual as compared to steady dates. Additionally, the number of refusals received was associated with the types of thoughts and actions commonly reported by perpetrators.
CONCLUSIONS: This simulation provides a new approach for examining sexual aggression in controlled experiments that vary factors within the simulation such as the man\'s and woman\'s intoxication and past sexual history. Because what happens is based on each individual\'s responses, it could be adapted for use in prevention and treatment programs.

Partnerships between academic and clinical-based health organizations are becoming increasingly important in improving health outcomes. Mutuality is recognized as a vital component of these partnerships. If partnerships are to achieve mutuality, there is a need to define what it means to partnering organizations. Few studies have described the elements contributing to mutuality, particularly in new relationships between academic and clinical partners. This study seeks to identify how mutuality is expressed and to explore potential proxy measures of mutuality for an alliance consisting of a hospital system and a School of Public Health. Key informant interviews were conducted with faculty and hospital representatives serving on the partnership steering committee. Key informants were asked about perceived events that led to the development of the Alliance; perceived goals, expectations, and outcomes; and current/future roles with the Alliance. Four proxy measures of mutuality for an academic-clinical partnership were identified: policy directives, community beneficence, procurement of human capital, and partnership longevity. Findings can inform the development of tools for assisting in strengthening relationships and ensuring stakeholders\' interests align with the mission and goal of the partnership by operationalizing elements necessary to evaluate the progress of the partnership.

The goals of this article are to review the major findings from alcohol administration studies that use sexual aggression proxies and to encourage additional experimental research that evaluates hypotheses about the role of alcohol in the etiology of men\'s sexual aggression. Experiments allow participants to be randomly assigned to drink conditions, therefore ensuring that any differences between drinkers and nondrinkers can be attributed to their alcohol consumption. One of the biggest challenges faced by experimental researchers is the identification of valid operationalizations of key constructs. The tension between internal and external validity is particularly problematic for violence researchers because they cannot allow participants to engage in the target behavior in the laboratory. The strengths and limitations associated with written vignettes, audiotapes, videotapes, and confederate proxies for sexual aggression are described. Suggestions are made for future research to broaden the generalizability of the findings from experimental research.