Patterns of Family Conflict and Accusations of Abuse in Dementia Family Caregivers: A Latent Class Analysis

Abstract

Background and Objectives

Family conflict is a stressor for dementia family caregivers, yet its impact may differ based on the relationship between caregivers and their recipients. This study’s objectives were to categorize caregivers into groups based on family conflict, examine whether the relationship to the recipient influences group membership, and determine whether these groups are associated with engaging in abusive and neglectful behaviors.

Research Design and Methods

This national, cross-sectional study of 453 dementia family caregivers used latent class analysis to generate groups based on family conflict and abuse accusations. A multinomial logistic regression determined if relationship type (i.e., being a spouse, child, or grandchild to the care recipient or having a nontraditional relationship) predicted group membership. Groups were examined as predictors of abusive and neglectful behaviors using analysis of variance.

Results

A 4-class solution emerged as the best fit: 3 groups with varying probabilities of family conflict and 1 group with elevated probabilities of abuse and neglect accusations. Relationship typed predicted membership in these classes. Group membership predicted abusive and neglectful behaviors.

Discussion and Implications

Adult children were more likely to experience high amounts of family conflict, whereas nontraditional caregivers were less likely to experience abuse and neglect accusations. Membership in the accusations group was a unique risk factor for abusive and neglectful behaviors. These findings support the need for continued investigation of family conflict in dementia caregivers. They also call to examine how family relationship types, including nontraditional family structures, influence caregiving outcomes.

Dementia is estimated to affect 55 million people globally (Gauthier et al., 2022). The burden associated with caregiving often falls on family of people with dementia (Brodaty & Donkin, 2009). Family caregivers are estimated to provide approximately $272 billion worth of unpaid care annually (Gaugler et al., 2022). Caregiving can take a serious emotional toll on the family. Family caregivers are more likely to experience depression, anxiety, and worse general mental health than noncaregivers (Ma et al., 2018). Although several sources of psychological distress for family caregivers have been identified, such as stress related to managing behavioral symptoms of dementia, most research has not examined factors outside the dyadic caregiving role (see Ma et al., 2018, for a meta-analysis). However, caregiving dyads are part of larger families with their own dynamic ways of relating to each other built on a lifetime of interactions. Conflict with other members of the family may be an important source of stress affecting outcomes in the caregiving dyad (Dieker et al., 2019; Yun et al., 2021). In particular, family conflict may present a unique risk factor for abusive and neglectful behaviors, but this has yet to be evaluated extensively by prior research. The purpose of this study was to investigate whether family caregivers could be grouped by family conflict profiles, how relationships between caregivers and their family members with dementia impact conflict profiles, and whether care behaviors are affected by these profiles.

Conflict in Dementia Family Caregiving

The stress process model of caregiving is a framework that describes how primary stressors of family caregiving, such as behavioral symptoms of dementia and care requirements, affect caregivers’ stress appraisals (Judge et al., 2009; Pearlin et al., 1990; Son et al., 2007). These stressors negatively affect caregivers’ stress appraisals related to the caregiving role which further affects their mental health and caregiving behaviors. This model also describes secondary stressors, which stem from the primary stressors of family caregiving, such as conflicts within the family about care decisions, the amount of care provided by family members, and accusations of financial abuse and neglect among family members (Scharlach et al., 2006). In this context, family conflict centers around how the relative with dementia is being treated by the caregiver, what should be done about their care, financial responsibilities related to caring for the relative with dementia, whether other family members are doing enough to help with care responsibilities, and other disagreements related to care responsibilities (Goode et al., 1998; Haley et al., 1987; Pearlin et al., 1981; Son et al., 2007). Family conflict is linked to worsened stress appraisals and mental health in caregivers (Butterworth et al., 2010; Koerner & Kenyon, 2007; Kwak et al., 2012; Strawbridge & Wallhagen, 1991). Thus, it is important to study family conflict in the context of caregiving, as it is an important predictor of caregiver mental health, which affects care behaviors.

Not all caregivers experience conflict with other family members, and not all conflicts are the same. Dementia family caregivers have unique demands such as more intense care demands, along with the behavioral symptoms and memory issues of their relative with dementia, and this may contribute to additional conflict in the family particularly around differing opinions on what care is necessary, perceived limitations of the care recipient and equitable distribution of responsibilities. Adult children caring for a parent experience more family conflict than caregiving spouses, and family conflict is a predictor of caregiver burden for adult child caregivers but not spouses (Dieker et al., 2019; Yun et al., 2021). The conflict between adult children and their parents is a source of stress, and adversely affects their relationship quality (Birditt et al., 2009). Thus, conflict between family members is perceived differently based on their relationship, which may affect the amounts and types of conflicts that arise between caregivers and other family members.

Abusive and Neglectful Behaviors Among Family Caregivers

Approximately half of dementia family caregivers self-report engaging in abusive and neglectful behaviors toward their care recipient (Pickering et al., 2020; Wiglesworth et al., 2010). In a study comparing care recipients who had and had not experienced abusive and neglectful behaviors, researchers found those who had experienced these behaviors were more likely to belong to a family that experienced conflict prior to the dementia diagnosis (Sadler et al., 1995). Caregivers who experience patterns of child maltreatment and intergenerational violence, indicators of severe family conflict throughout the lifespan, are at greater risk of engaging in abusive and neglectful behaviors when they become caregivers (Dong & Wang, 2019; Herrenkohl et al., 2022; Kong & Easton, 2019). An Australian qualitative study examining applications for guardianship found that many of the family conflicts among siblings were accusations of neglect and financial mismanagement towards one “black sheep” sibling (Peisah et al., 2006). More work is needed to elucidate the effects of family conflicts and caregiver–recipient relationships specifically in dementia caregivers, as they are likely to experience unique challenges and conflicts in dementia caregiving.

It is necessary to determine whether unique profiles of conflict frequency may differentially affect the use of abusive and neglectful behaviors by caregivers. Researchers have measured family conflict as a continuous variable or as a simple count of family conflicts (Fukui et al., 2018; Semple, 1992). However, linear relationships between family conflict and abusive and neglectful behaviors may not address patterns of conflict among caregivers in reality. A person-centered approach where unique patterns of conflict frequency are identified may be beneficial in identifying both theoretically and clinically relevant risk groups for abusive and neglectful behaviors based on family conflict patterns. Identifying risk groups may inform future interventions targeting these risk groups.

Current Study

The first objective of this study was to determine whether dementia caregivers could be categorized into unique, clinically relevant groups based on family conflict profiles. Based on research grouping caregivers based on stressors and buffers against caregiving stress including dementia severity, social support, and family function, we hypothesized that caregivers could be categorized into three unique groups based on their family conflict and accusations of abusive behaviors toward the care recipient: low, moderate, and high conflict (Graven et al., 2023; Yuan et al., 2020). In this context, accusations are direct confrontations of abuse of the care recipient from others toward the caregiver or from the caregiver toward others. Due to the novelty of investigating accusations in this way, no hypotheses were advanced regarding the distribution of accusations among conflict groups. The second objective was to determine caregiver–recipient relationship (spouse, child–parent, grandchild–parent, and nontraditional caregivers) influenced group membership. We hypothesized that spouses would be more likely to be categorized in a “low conflict” group (Dieker et al., 2019; Yun et al., 2021). We advanced no hypotheses regarding other relationships. Our third objective was to determine whether relationship type and group membership were associated with the frequency of abusive and neglectful behaviors in the past 6 months. We hypothesized that spouses and individuals in the “low conflict” group would engage in abusive and neglectful behaviors less frequently than other groups.

Method

Participants

This was a cross-sectional, secondary analysis of baseline data from a national, multitime series longitudinal study on caregiving experiences. We included 453 adults over 18 years of age, who coresided with and provided unpaid care to a family member at least 60 years old with mild cognitive impairment or dementia based on their score on the AD8 (Galvin et al., 2005). Participants were excluded if they did not have reliable internet access or could not speak and read English or Spanish. Participants were recruited nationally through social media, online and print media, and community outreach based on suggested best practices to protect the validity of the data in online research (Kramer et al., 2014; Tarzia et al., 2017; Teitcher et al., 2015).

Data Collection

Data collection began after the participant consented to the research in the online eligibility survey. A link to the online baseline survey was sent to the email address provided by the participant. Email, text, and phone reminders were sent (depending on participant preference indicated during the eligibility survey) to encourage survey completion. Participants were compensated with a $40 Amazon e-gift card for survey completion.

Measures

Demographics were collected for each participant and included caregivers’ age, household income, race/ethnicity, and gender, along with recipients’ Eight-item Informant Interview to Differentiate Aging and Dementia (AD8; Galvin et al., 2005). Descriptive statistics for the entire sample divided by relationship type are provided in Table 1.

Family conflict

Family conflict was measured using the Family Conflict Scale (FCS; eight items) and four investigator-created questions about elder abuse accusations between the caregiver and others (Gaugler et al., 1999; Semple, 1992). The FCS assesses how much disagreement (0, “no disagreement” to 3, “quite a bit”) caregivers have with family members outside the dyad in the past 6 months about the caregiving workloads, respect toward the caregiver, and treatment of the care recipient. The total score is the sum of scores across the eight individual items. Scores range from 0 to 24, with higher scores indicating more conflict with other family members (Gaugler et al., 1999). Accusations of abusive and neglectful behaviors, another type of family conflict, were measured by asking participants whether or not (1) they accused others of committing financial abuse or (2) elder abuse/neglect toward the care recipient, and whether others accused them of those same actions. The total score is a count of “yes” responses. Scores range from 0 to 4. Table 2 provides counts for individual conflict items.

Relationship type

Participants were asked about their relationship with the care recipient with dementia. Based upon inclusion criteria responses included opposite-sex husband/wife/spouse, opposite-sex unmarried partner, same-sex husband/wife/spouse, same-sex unmarried partner, parent, stepparent, parent-in-law, grandparent, step-grandparent, and grandparent-in-law.

Elder abuse and neglect

Participants also answered a series of nine questions that assessed the frequency with which they engaged in self-reported abusive and neglectful behaviors toward the care recipient in the last 6 months. We collapsed items within subscales from the Conflict Tactics Scale—Revised (CTS2) and Conflict Tactics Scale—Older Adults (Beach et al., 2010; Pickering et al., 2020; Straus & Gelles, 1987; Straus et al., 1996). Consistent with CTS2 scoring, the response options for these questions were coded: 0 (never), 1 time, 2 times, 3–5 times (3), 6–10 times (4), 11–20 times (5), and more than 20 times in the past 6 months (6). Responses were recoded to the midpoint. Thus, participants’ frequencies for individual abusive and neglectful behaviors were scored as follows: 0 = 0, 1 = 1, 2 = 2, 3 = 4, 4 = 8, 5 = 15, and 6 = 25 (Straus & Gelles, 1987). A final, continuous, abusive, and neglectful behaviors score was calculated as a sum of these frequencies.

Evaluation of Missing Data and Imputation Procedures

Data were missing or not reported for relationship type (N = 14, 3.09%), age (N = 7, 1.55%), and gender of the caregiver (N = 2, 0.44%). Imputation was not performed to replace relationship type, caregiver age, recipient AD8, household income, caregiver race/ethnicity, or caregiver gender because these are inherent characteristics of the caregiver. Because of the importance of relationship type to these analyses and to ensure interpretability of intercepts in the model, a variable for “missing relationship type” was dummy coded, thus each participant had a yes/no variable for every possible relationship type. For other variables, we used a two-level imputation procedure to replace missingness (Asparouhov & Muthén, 2010). This was done 40 times, and one imputed data set was selected randomly for analysis. Imputation procedures were performed in Mplus 8.8 (Muthén & Muthén, 1998-2017).

Data Analysis

A three-step latent class analysis (LCA) was conducted to identify latent classes among participants. The three-step process first estimates latent class parameters based on the theoretically connected indicator variables and a specified number of classes, then classifies participants based on the classes defined in the first step, and finally, performs a multinomial logistic regression to examine the effect of covariates on class membership (Vermunt, 2010). The benefit of performing these three steps as one analysis is to account for the underestimation of the effects of covariates on class membership (Vermunt, 2010). First, latent classes were estimated based on the 12 family conflict indicator variables. We recoded the FCS items to dichotomous variables, with answers “0, no disagreement” and “1, rarely disagreement” coded to “1—low conflict,” and answers “2, some disagreement” and “3, quite a bit of disagreement” coded to “2—high conflict.” This recoding was based on the distribution of answers for these items to support the interpretability of results. Given the limited research in this area, we did not have a priori classifications of family conflict patterns. LCA was used to identify unobserved categories based on family conflict patterns. Thus, our classifications are empirically derived based on the latent characteristics. We based the number of classes in the final model based on fit statistics, not a priori fit criteria. We iterated on the models by increasing the number of possible classes and determined the best solution based on the model fit indices, including comparisons of Akaike Information Criteria and adjusted Bayesian Information Criteria, along with Lo–Mendell–Rubin and Bootstrap Likelihood Ratio tests for model fit (Morovati, 2014). Additionally, entropy values were examined (Celeux & Soromenho, 1996). Entropy is a value between 0 and 1 that measures how well the model identifies unique classes. Values over 0.8 are considered acceptable, with values closer to 1 being preferable (Celeux & Soromenho, 1996; Weller et al., 2020). Posterior probabilities, or the probability that a participant is classified in a given class, were generated for each participant for each class. These probabilities ranged from 0 to 1 with 1 being the highest. Then participants’ highest posterior probabilities for class membership were used to create a latent class variable.

The analysis included a multinomial logistic regression with the latent class membership variable as the outcome and relationship types as independent variables (Asparouhov & Muthén, 2014). To improve the interpretability of regression coefficients and to create adequate cell sizes for relationship type, participants were dummy coded (assigned a yes/no variable) into five total categories: spouses, children, grandchildren, nontraditional caregivers consisting of individuals taking care of someone that was neither their spouse nor biologically related to them (step-family and in-laws in this sample), and individuals who were missing relationship type. R3STEP in MPlus 8.8 was used to perform the automatic three-step procedure for LCA (Muthén & Muthén, 1998-2017).

Finally, we used relationship type and class membership to predict the frequency of abusive and neglectful behaviors in the past 6 months reported by caregivers using linear regression and one-way ANOVA in IBM SPSS 28. Tukey’s correction for multiple comparisons was used to limit the possibility of type I errors.

Results

Descriptive Statistics

Participants (N = 453) were 52.56 years old on average, 394 (87.4%) were female, and 232 (51.2%) were non-Hispanic White. In relation to the care recipient, 101 (22.3%) caregivers were spouses, 239 (52.8%) were children, 53 (11.7%) were grandchildren, and 46 (10.2%) were nontraditional caregivers (stepchildren, step-grandchildren, and in-laws). Two hundred and fifteen (47.5%) of participants reported at least one abusive or neglectful behavior in the last 6 months.

Objective 1: Identify Latent Classes

The number of caregivers who reported high amounts of each individual conflict ranged between 28.9% and 44.8%. Between 3.1% and 6.6% of caregivers reported individual accusation items. A four-class solution was selected based on multiple criteria (see Table 3 for various model fit statistics). Model fit statistics favored a five-class solution; however, classes created with the five-class and six-class solutions were not as distinct from each other as the four-class solution and were less interpretable due to small cell sizes (multiple odds ratios and standard errors in the third step became undefined). The entropy value for the four-class solution was 0.90, indicating acceptable classification quality. Further, the four-class solution was chosen as the classes that emerged were more theoretically grounded and clinically meaningful, which was key to this study’s person-centered approach to finding groups who were at risk for abusive and neglectful behaviors.

In total, the four classes represent a low amount of family conflict (low), a moderate amount of family conflict (moderate), a high amount of family conflict (high), and a distinct fourth group that experiences moderate–high amounts of accusations in addition to moderate amounts of conflict (accusations). All three of the other groups had a low number of accusations. Figure 1 depicts posterior probabilities of endorsing the 12 indicator variables for the four identified classes: The “low” group (n = 229, 50.0%) had low (0.00–0.06) probabilities of endorsing all items, indicating that this class had minimal family conflict and accusations. In the “moderate” group (n = 129, 28.5%), probabilities for the eight FCS items ranged from 0.5 to 0.8 with low (0.00–0.03) probabilities for the “accusations” items. The “high” group (n = 80, 17.7%) had high (0.8–1.0) probabilities of endorsing all FCS items, whereas the probabilities of the “accusations” items were 0.10–0.15. The “Accusations” group (n = 15, 3.8%) had moderate (0.29–0.61) probabilities of FCS items, and moderate–high (0.31–0.84) probabilities of “accusations” items.

Objective 2: Does Relationship Type Influence Group Membership?

We predicted latent class membership as an observed outcome using dummy-coded covariates for spouses, children, grandchildren, and nontraditional relatives. “Child” was set as the intercept group because it was the most common relationship type. As indicated in Figure 2, there were significant associations between relationship types and family conflict group classification. Spouses (B = 1.229, 95% confidence interval [CI] = 0.314–2.144), grandchildren (B = 1.210, 95% CI = 0.085–2.335), and nontraditional caregivers (B = 1.187, 95% CI = 0.016–2.359) were more likely (compared to children) to be classified in the “Low” group than the “High” group. Nontraditional caregivers were less likely to be in the “accusations” group versus the “moderate” (B = −3.171, 95% CI = −5.662 to −0.679) and “low” (B = −2.922, 95% CI = −5.258 to −0.586) conflict groups (compared to children; see Supplementary Table 1 for a full breakdown of multinomial logistic regression results).

Objective 3: Does Group Membership or Relationship Type Affect Abusive and Neglectful Behaviors?

To test whether class membership is associated with rates of abusive and neglectful behaviors in the past 6 months, we entered class membership as an independent variable in a one-way ANOVA, with abusive and neglectful behaviors as the dependent variable. The omnibus test revealed significant differences among classes (F(3, 449) = 7.04, p < .001). A post-hoc Tukey’s test for multiple comparisons revealed the “accusations” class had significantly more abusive and neglectful behaviors on average compared to each other class (all p’s < .05). Additionally, the “high” class had significantly more abusive and neglectful behaviors on average than the “low” class. A similar analysis was performed with relationship type as the independent variable. Relationship type was not associated with abusive and neglectful behaviors (p = .210). Results for these analyses are provided in Table 4.

Discussion

This study fills a gap in the literature by identifying unique and meaningful categories of caregivers based on their family conflict profiles. To our knowledge, this is the first report of latent classes to categorize caregivers by the level of family conflict and accusations of abuse and neglect. Our first hypothesis was supported, as unique groups of caregivers emerged. However, three were characterized by varying levels of family conflict with relatively lower levels of accusations, whereas there was a fourth group characterized by high levels of accusations of abuse and neglect among family members. Of note, the accusations group had lower levels of family conflict relative to the level of accusations and also had lower levels of family conflict relative to the high conflict group. Or conversely, the groups characterized by conflict reported few, if any, accusations among them. It is not simply disagreement and tension among family members that is associated with abuse and neglect—these accusations seem to be distinct from general family conflict. These findings suggest a need for more work examining the relevance and impact of accusations among caregivers. Because these are associated with abuse, they may present a better screening option than actual abuse.

Regarding our second hypothesis, relationship type was a predictor of group membership and spouses were classified in the low conflict group. This study was novel in that we investigated family conflict and elder abuse and neglect in nontraditional caregivers. Nontraditional caregivers are not frequently targeted by family caregiving studies and interventions, but in this study, we found that they represented a sizable portion of our research participants (~10%). Nontraditional caregiving is on the rise, and one reason for this is an increase in the number of older adults getting remarried, thus having stepchildren who later become their caregivers (Lewis & Kreider, 2015; Sherman et al., 2013). Our finding that nontraditional caregivers were less likely to fall into the high conflict and accusations groups is in contrast to findings by other researchers that found stepchildren are more likely to have disagreements with other family members about finances and care decisions (Sherman et al., 2013). A potential reason for this contrast is that our nontraditional caregiver group included step-grandchildren and in-laws as well as stepchildren. Interestingly, although nontraditional caregivers were less likely to fall into the high conflict and accusations groups compared to children, they did not report fewer abusive and neglectful behaviors. This implies that nontraditional caregivers may experience unique challenges and conflicts that were not assessed here, such as financial concerns, issues related to family dynamics of nontraditional families, and additional accusations. Future work should identify unique challenges and conflicts by specifically recruiting and interviewing nontraditional caregivers.

We included grandchild caregivers in our analyses on relationship type. There are not reliable estimates of the prevalence of biological grandchildren providing care for grandparents with dementia, but it is thought that 4%–10% of caregivers are biological grandchildren (Venters & Jones, 2021). Similar to nontraditional caregivers, grandchildren represented just over 10% of our sample. Grandchildren were less likely to fall into the high conflict and accusations classes than children, and although the omnibus ANOVA comparing relationship types on the frequency of abusive and neglectful behaviors was not significant, grandchildren reported less than half the number of these behaviors compared to children. One reason for these findings could be that grandchild caregivers are younger than child caregivers, on average. Younger caregivers spend less time caring on average, may view the experience of caregiving for a grandparent more positively, and may also have greater resources for dealing with the caregiving role (D’Amen et al., 2021; Pristavec, 2019). These resources include having more help from other caregivers and having greater access to online resources (including social media and health-related information; Kim, 2015; Patrick et al., 2023). In line with the stress process model, these resources and benefits of caregiving are likely to result in better stress appraisals, which, in turn, are associated with fewer abusive and neglectful behaviors. Further investigation is needed to understand the mechanisms behind being a grandchild caregiver as a protective factor against abusive and neglectful behaviors.

Finally, our third hypothesis was partially supported. Caregivers in the high conflict and accusations groups reported more abusive and neglectful behaviors than other groups. A key finding is that the accusations group also did not experience high amounts of conflict. The Adult Maltreatment Report in 2021 found that 28.7% of Adult Protective Services (APS) claims were substantiated (meaning enough evidence of abuse or neglect was present to take further action; McGee & Urban, 2022). Additionally, family members reported 16% of all APS claims in 2021. Providers and social workers report fear of alienating patients and discomfort with direct abuse and neglect questioning as barriers to routine screening for elder abuse and neglect in practice (Schmeidel et al., 2012). Questioning caregivers about experiences of conflict with other family members, including whether there have been any accusations, maybe a good way of transitioning into routine screening for elder abuse and neglect and identifying those caregivers who need more thorough assessment. Future research should investigate the mechanisms by which family conflict leads to more frequent acts of abusive and neglectful behaviors, such as the nature, direction, source, and proximity of the conflict, and whether high levels of family conflict can predict the onset of abuse and neglect to inform the development of interventions.

One limitation of this study is that accusations of elder abuse/neglect were reported very infrequently. Thus, the group that experienced more frequent accusations was relatively small. This prevented us from being able to look at the influences of each of the nine relationship types originally collected on family conflict classification due to small comparison groups. Additionally, we did not measure financial exploitation due to difficulty in judging the validity of financial exploitation given varying laws guiding this (e.g., powers of attorney, etc.) in the nationally recruited sample. Because finances are an important source of conflict, this is an area for future research.

The FCS addresses many of the conflicts dementia caregivers experience; however, this was not an exhaustive assessment of conflicts dementia caregivers experience. Future work should investigate conflict unique to dementia caregivers. Additionally, this study focused on cross-sectional data, which precludes the examination of the dynamic nature of family conflict. Future studies could benefit by examining conflict longitudinally or on a daily level. Such examinations would help identify how conflict group membership may change and how conflict influences caregiver behaviors over time.

In summary, this study contributed to the growing body of literature on the stress process model of caregiving by describing a secondary stressor, family conflict, and classifying participants based on their conflict group probabilities (Pearlin et al., 1981, 1990). The latent classes identified here are unique groups of dementia caregivers who experience different amounts and types of conflicts with and accusations among one another. These findings indicate that it is important to employ a person-centered approach when investigating family conflict among caregivers, as an examination of linear relationships would not have identified these unique groups. This study was novel in that we provided evidence that nontraditional and grandchild caregivers have unique needs compared to child caregivers. Researchers should target these caregivers for studies that identify these unique needs. Future work should also examine how family conflicts vary daily, and how fluctuations in these conflicts might present unique challenges for caregivers. Finally, individuals who provide services to individuals with dementia and their caregivers should consider the social situation outside the care dyad, as this may affect care outcomes and behaviors by the caregiver. Family conflict and accusations should be screened to detect potential abusive and neglectful behaviors, as this may be more beneficial than screening all caregivers for these behaviors.

Supplementary Material

Supplementary data are available at The Gerontologist online.

Funding

This work was funded by the National Institute on Aging (grant R01AG060083).

Conflict of Interest

None.

Data Availability

This data will be made available in the national archive of computerized data on aging upon completion of the NIH award. This study was not preregistered.

Author Contributions

W. Browning—Conceptualization, data curation, formal analysis, methodology, software, validation, visualization, writing—original draft, and writing—review and editing. M. Yildiz—Data curation, formal analysis, methodology, software, validation, and writing—review and editing. C. Maxwell—Methodology, validation, writing—review and editing. T. Sullivan—writing—review and editing. M. Yefimova—Writing—review and editing. J. A. Hernandez Chilatra—Writing—review and editing. C. Pickering—Conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, writing—review and editing

References