Detection and management of suspected infections in people with dementia – A scoping review of current practices

People with dementia have an increased risk of hospitalization and mortality due to infections. We aimed to explore decision-making processes and interventions for detecting and managing suspected infections in people with dementia and involved actors and determinants. We conducted a scoping review, searching CINAHL and PubMed, and synthesized data through mapping and narratively. We identified 22 studies, based mostly on nursing homes and US data. Decision-making processes included recognition of infections based on observations of early signs and symptoms, actions when suspecting infections, and proxy/family involvement. Interventions included antimicrobial stewardship and other decision-support tools. Determinants included healthcare staff perceptions, and other system/person-related factors. Healthcare staff were the main actors, proxy/family were mentioned scarcely, and people with dementia only once. Our findings show scarcity of evidence on people with dementia and outside of the nursing homes. We highlight knowledge gaps and inform research shaping interventions for improving infection detection and management.


Introduction
Global forecasts paint a concerning picture: the number of People with Dementia (PwD) is predicted to increase threefold by 2050 (Nichols et al., 2022).PwD live with more comorbidities, have higher healthcare expenses, and lower quality of life than people without dementia (Nandi et al., 2022;Velandia et al., 2022).Infections are a frequent comorbidity in PwD (Scrutton and Brancati, 2016).Not only are PwD at a higher risk of infections, (Asby et al., 2021;Ecarnot et al., 2023) but infections may also be misdiagnosed and inadequately treated due to difficulties in adhering to care plans, (Daiello et al., 2014;Livingston et al., 2020) atypical clinical presentation of infections, (Scrutton and Brancati, 2016;Tingström et al., 2015;Kovach et al., 2010;van der Maaden et al., 2015;Komagamine et al., 2022) eating and drinking difficulties, (Livingston et al., 2020;Payne and Morley, 2018) and a second-hand clinical history due to trouble communicating symptoms (Scrutton and Brancati, 2016;Daiello et al., 2014;Livingston et al., 2020;Kovach et al., 2010;van der Maaden et al., 2015).Indeed, PwD have been shown to be at a markedly higher risk of hospital admission with infection (Janbek et al., 2021a), readmission, (Janbek et al., 2021b) and excess mortality due to infection, (Janbek et al., 2021c) compared to people without dementia.Given the ineffectiveness of existing interventions to decrease hospitalizations (any reason, including infections), (Livingston et al., 2020;Phelan et al., 2015;Naylor et al., 1999) and the highlighted complexity in diagnosing and managing infections in PwD, there is a need for effective evidence-based complex interventions for diagnosing and managing infections in PwD (Livingston et al., 2020;Phelan et al., 2015).Such interventions first require understanding the practices involved in the detection and management of infections.This includes understanding actors' needs and challenges, how they are brought into the interventions, and how the interventions are created (Skivington et al., 2021) and evaluated in sustainable ways (May and Finch, 2009).Further, understanding current practices involved in detecting infections in PwD is an important step to provide clarity over the best way to deliver care (Phelan et al., 2015).However, past research on this topic is fragmented and narrowly focused, (van der Maaden et al., 2015;Hendriks et al., 2017;Yates et al., 2015;Mitchell et al., 2014;Parsons and van der Steen, 2017) and most research on early infection detection and decision-supporting tools for older adults (Masot et al., 2022) did not include PwD.Therefore, a comprehensive overview of infection detection practices and management is lacking.This highlights the need for the current review, which contributes to developing intervention objectives based on the needs assessment (Bartholomew et al., 1998) by gathering knowledge on practices involved in infection detection and management including decision-making processes, interventions targeting these processes, actors involved, and their determinants.We thus aimed to: 1) Identify and synthesize literature on healthcare decision-making processes in primary and secondary care, involved actors, and considered determinants for detecting and managing suspected infections in PwD. 2) Identify existing interventions targeting decision-making processes, actors, and determinants.

Methods
This is a scoping review, employing the methodological framework developed by Arksey and O'Malley (Arksey and O'Malley, 2005;Levac et al., 2010).The research team developed an a-priori protocol (Supplement A).Fig. 1 provides an overview of the research questions and scope.

Inclusion criteria
We used The Population, Concept, and Context format (Pollock et al., 2023) to present our inclusion criteria to identify and select relevant studies.

Population
This review targeted studies that included people with any dementia type/stage suspected of having any infection type.If the study's population was majority PwD with suspected infections, then the study was fully included.For studies that lacked explicit details about the study population but included PwD with suspected infections, or where infections in PwD were not the primary focus but were nonetheless addressed, only the relevant sections referred to as snippets were included.

Concept
We targeted studies addressing any of the four main concepts, defined in the following.
Decision-making processes: Steps leading to detection and management of suspected infections, including considerations, gathering information, making decisions, and assessing solutions.
Interventions targeting the decision-making processes: Developed actions and strategies to improve the detection and management of suspected infections, including tools, algorithms, and care plans.Actors: People taking an active part in the infection detection and management practices, including PwD, their family/proxies, and healthcare staff.
Determinants: Needs, attitudes, experiences, and perspectives of actors including challenges, facilitators, and patient, social, and healthcare system-related factors.

Context
All countries and healthcare settings were included.We excluded studies on developed interventions without implementation or impact data, and articles that focused strictly on end-of-life or palliative care (Supplement A lists excluded studies through full-text review and reasons for exclusions).

Search strategy
In March 2023, two authors (MI and JJ) searched PubMed and CINAHL databases for dementia, infection, detection/management, and synonyms thereof (search terms detailed in the Protocol, Supplement A) to find original peer-reviewed studies published in English between January 2010 and March 2023.Studies were searched for and selected in an iterative process (Arksey and O'Malley, 2005).

Data extraction
Two reviewers (MI and JJ) independently performed data extraction using a developed pilot-tested and review-specific form to ensure completeness and systematization (presented in the Protocol, Supplement A).For snippets, only the relevant findings were of interest and thus extracted.Disagreements in data extraction were resolved through discussion until consensus.The PRISMA extension for scoping reviews (Tricco et al., 2018) (Supplement A) guided our reporting.

Critical appraisal of studies
Critical appraisal was conducted for all studies except snippets.Risk of bias tools used were the Newcastle Ottawa Scale for cohort studies (modified, see Supplement B), (Lo et al., 2014) The Revised Cochrane Risk of Bias Tool for Cluster Randomized Trials, (Eldridge et al., 2021) The Critical Appraisal Skills Program for the qualitative study, (CASP Critical Appraisal Checklists, 2023) the Mixed Method Appraisal Tool for the mixed method study, (Hong et al., 2018) and a Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies (Whiting et al., 2011) for the study using a machine learning diagnostic method (Enshaeifar et al., 2019).Evaluation conducted using a Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies consisted of assessing each domain for risk of bias and for the first three domains, concerns regarding applicability (Whiting et al., 2011).MI and JJ conducted all appraisals separately and discussed results until consensus.

Data synthesis
We synthesized findings through concept mapping (Trochim, 1989) for a graphical presentation and narratively (Popay et al., 2006) (involved comparing findings to identify similarities, differences, or patterns in described practices).We grouped findings under themes which covered different aspects of decision-making processes and interventions.We did not differentiate between fully included studies and snippets during synthesis.
We did not identify any studies in the secondary healthcare setting.All but one study (Enshaeifar et al., 2019) were conducted in NHs/equivalent care settings.Healthcare staff (actors) were investigated in all the studies, with only five studies that additionally investigated proxies/families of PwD as actors, (Hendricksen et al., 2022;Mitchell et al., 2021;Juthani-Mehta et al., 2009;Hartman et al., 2023;Givens et al., 2009) and one study that investigated PwD as actors (Juthani-Mehta et al., 2009) (Figs. 3 and 4).

Critical appraisal of studies
Two cohort studies (Yates et al., 2015;Hendricksen et al., 2022) had low risk of bias, two moderate, (Tingström et al., 2015;Givens et al., 2015) and one high (Boockvar et al., 2000).One randomized controlled trial showed high risk of bias, (Mitchell et al., 2021) and the machine learning study showed low concerns regarding applicability but high risk of bias (Enshaeifar et al., 2019).A qualitative study had a low risk of bias (Tingström et al., 2010).A mixed method study had high quality in the qualitative section, but highlighted areas of concern within the quantitative and mixed-methods components (Allemann and Sund-Levander, 2015).Detailed assessments are in the Supplement B.

Data synthesis
Table 2 presents a summary of study results.Findings under themes for decision-making processes and interventions are graphically presented in Figs. 3 and 4, and narratively presented in the following.

Recognition of infections based on early signs and symptoms
3.2.1.1.1.Observation: Physical signs and symptoms.Nursing assistants observed NH resident/PwD "seems to be ill" based on general signs and symptoms (mostly indicating fever, e.g., warm or changed color of the skin), specific signs and symptoms (e.g., breathing difficulties), and pain (Tingström et al., 2010).They perceived this observational category as a more distinct infection indicator than "not as usual" (observed during everyday interactions), and majority of them perceived identifying infections generally difficult.Nurses and physicians placed bigger emphasis on specific signs and symptoms, compared to the general ones, based on nursing assistants´experience.Moreover, nursing assistants perceived that, unlike them, nurses and physicians considered that temperatures below 38 • C do not indicate fever (Tingström et al., 2010).In another study (Kovach et al., 2010) temperature was measured for 61 % PwD who developed infections (41 % of which had >37⋅2 0 C).Moreover, when observing clinical reasons for urinary tract infection    (Tingström et al., 2010) observed NH residents/PwD being "not as usual" based on behavioral changes, discomfort, confusion, and other.They perceived it challenging to sort out behavioral changes due to NH resident/PwD differing from one another in normal appearance or behavior, and in infection-indicating behavior, (Tingström et al., 2010) and they perceived it challenging to sort out behavioral changes if they did not know the PwD for some time (Sund-Levander and Tingström, 2013).In another study, (Kovach et al., 2010) nurses found several behavioral changes to be associated with new infections, namely resistive behavior (in 47 % of PwD with infections), calling help or something is wrong (39 %), and a distressed facial expression (25 %).Focus groups including nurses, other healthcare assistants, and care home managers observed behavioral changes and confusion as common infection symptoms, (Jones et al., 2020) and in a different study, (Hartman et al., 2022) nurses identified confusion and deterioration as infection symptoms.When observing clinical reasons (other than physical signs) for UTI suspicion, (Juthani-Mehta et al., 2009) nurses and physicians listed mental status and behavior changes as most common reasons (in 39 %, and 19 % of 399 episodes, respectively).

Diagnostic certainty.
A study on physicians diagnosing pneumonia (Helton et al., 2011) showed that those who ordered a chest x-ray (vs.those who did not) had lower NH presence (defined as the percentage of working hours conducted in NH, and the frequency of visits to PwD; OR: 1⋅9 p<0⋅05).However, physicians who were certain of the diagnosis and those who were not had similar presence in NH.

Actions when suspecting infections
3.2.1.2.1.Conducting procedures.One study showed that majority of PwD underwent at least one procedure conducted by physicians, physician assistants, and nurses, with average use of 3⋅7 procedures/ person-year (Yates et al., 2015).Most frequent procedures were blood draws and urinalysis, with the highest procedure utilization rates for suspected UTI, followed by respiratory tract infection (RTI).Key determinants for greater procedure burden (based on the number and type of procedures received) included black race, lack of do-not-hospitalize order, not being enrolled in hospice, and suspected infection source being UTI, RTI, and fever of unknown source, compared to skin/soft tissue infections.Several PwD characteristics such as comorbid conditions, and proxy visits did not significantly influence procedure burden.
3.2.1.2.2.Decisions to contact physicians.When nurses at the NH were presented with 4 choice scenarios (2 with dementia and 2 without), their decision to call a physician about the suspected infection was not influenced by dementia (Beeber et al., 2021).Moreover, nurses emphasized the need to wait with contacting a physician when based only on PwD emotional change (Hartman et al., 2022).
3.2.1.2.3.Antimicrobial prescribing.Nurses emphasized the need to wait with prescribing when based only on PwD emotional change (Hartman et al., 2022).Opinions about using antimicrobials for non-specific symptoms varied among physicians (Hartman et al., 2022) who, together with nurses, perceived challenges in treatment due to PwD non-specific symptoms, poor medical history, and high prevalence of asymptomatic bacteriuria (Harbin et al., 2022).Moreover, physicians and nurses perceived PwD voice regarding antimicrobial prescribing important, even when consent incompetent (Harbin et al., 2022).When physicians at the NH were presented with 4 choice scenarios (2 with dementia and 2 without), their decision to prescribe antibiotics about the suspected infection was not influenced by dementia (Kistler et al., 2020).
3.2.1.2.4.Urinalysis.Another action was urinalysis, which physicians, (Hartman et al., 2022) and focus groups of nurses, other healthcare assistants, and care home managers (Jones et al., 2020) perceived important, particularly in infection uncertainty and when combined with PwD examination, (Hartman et al., 2022) but they also perceived it challenging to obtain, particularly among incontinent (Jones et al., 2020).
3.2.1.2.5.Other actions.One study identified 11 actions by nurses and nursing assistants when suspecting an infection, including observation as the most frequent, followed by contacting physician (Allemann and Sund-Levander, 2015).The action nothing happened was present in almost half of the suspected infection episodes and more commonly when episodes were initiated by nursing assistants.

Proxy/family involvement
3.2.1.3.1.Awareness of suspected infections and documented discussion about them.Proxies to PwD in NHs (majority children of PwD) were aware of only 40 % of suspected infection episodes, and only around half of those were registered with a documented discussion with a physician, physician assistant, nursing practitioner, or nurse regarding infection episodes (Givens et al., 2015).There were no regulations on documenting these discussions in the NH under study, and informal discussions may have appeared undocumented.Overall participation of the proxies in treatment decision-making for the infection episodes was in more than half of episodes they were aware of.For decision-making about infection treatment in which they were not involved, only 15 % wished to be involved.Key determinants for higher proxy awareness included more antimicrobial prescriptions and more hospital transfers.Determinants associated with more documented discussions included PwD receiving care in dementia special care unit, receiving more visits, and not being on hospice.PwD and proxy factors such as age or race showed no significant associations.Nurses were most involved in documented discussions (56 %), followed by nurse practitioners (33⋅8 %), and physicians (13 %).
A trial (Mitchell et al., 2021) assessed the impact of a complex intervention for nurses, physicians, physician assistants, nurse practitioners, and proxies on the number of antimicrobial courses and procedures.The outcome was 33 % less prescribed antimicrobials for PwD (non-significant difference compared to routine care).In a follow-up

Table 2
Summary of the study results.

Study
Theme and findings Actors, determinants, and findings

Decision-making processes
Fully included (Tingström et al., 2010) Recognition of infections based on observation of early signs and symptoms -observation of physical signs and symptoms, and observation of behavioral and mental status changes Two categories generated from described early signs and symptoms: -Is not as usual, infection indicators: e.g., changes of behavior, discomfort (facial expression), agitation, confusion/cognitive changes, tiredness, reduced eating, looking like physical and mental exhaustion -Seems to be ill, infection indicators: specific and general symptoms (e.g., panting, respiratory difficulties, bad-smelling and thick urine, fever, skin color changes, gastrointestinal and urinary issues, cyanosis, brief unconsciousness periods), pain Actors: Healthcare staff (nursing assistants) -Nursing assistants perceive challenges in identifying behavioral changes due to diverse normal behaviors and appearances in NH residents/PwD -13 of 21 nursing assistants perceived detecting suspected infections somewhat difficult, 4 somewhat easy, 1 hard, 1 easy, 0 no problem -Clearer infection indicators perceived in seems to be ill category, including distinctive physical signs, incontinence, or respiratory noises -Nurses and physicians focus on specific signs and symptoms, as perceived by nursing assistants -Difference in fever definition: nursing assistants observe fever signs at temperatures below and nurses and physicians above 38   -Difficulty recognizing UTI due to nonspecific symptoms: PwD not articulating symptoms clearly, atypical signs masked by other health issues e.g., incontinence or comorbidities, overlap between dementia and UTI symptoms, initial symptoms often limited to confusion or deterioration, leading to less distinct symptoms -Varied physician approaches to prescribing antibiotics for non-specific symptoms -Physical meetings and urine tests often necessary for accurate diagnosis, urinalysis essential in unclear symptom cases, especially in PwD with unexplained confusion, importance of positive test results in guiding further actions -Attitude to wait: not all unfavorable PwD days urge medication/calling a physician (Jones et al., 2020) Recognition of infections based on observation of early signs and symptoms -observation of behavioral and mental status changes Actions when suspecting infections -Urinalysis -Disorientation and behavioral changes identified as common infection signs -Suspected UTI cases usually lead to attempts to obtain a urine specimen Actors: Healthcare staff (nurses, other healthcare assistants, care home managers) Perceived challenges in collecting urine specimens especially with the presence of incontinence
-Reproducibility: Morning vs afternoon responses had 76 % agreement -Convergent validity: all items correlated with global status change; confused was endorsed 55 % of the time for PwD vs. 15 % for nondemented.
Actors: Healthcare staff (nurses, nursing assistants, physicians) -Nursing assistants' daily instrument completion: 15 minutes or less for all assigned NH residents/PwD, not completing instrument in 42 % of work shifts primarily due to lack of time -Majority perceived tool as potential enhancer of communication with other healthcare staff (Tingström et al., 2015) Early Detection Scale of Infection tool (EDIS) Looks at physical signs and behavioral changes.Outcomes: -Content validity: 12/13 items correlated significantly with at least one other -Construct validity: items temperature, respiratory symptoms, and general signs and symptoms of illness were significantly linked to infection -Changes in signs and symptoms and verified Infection: accurately predicted responses in 61 % of 59 cases analyzed, prediction range: 0-84 % -Specificity and sensitivity: respiratory symptoms 51 % and 29 %, general signs and symptoms of illness 41 % and 30 % Actors: Healthcare staff (nursing assistants) (Enshaeifar et al., 2019) Machine Learning tool Algorithm consisted of supervised and unsupervised models for detecting UTI.analysis of the trial, the authors (Hendricksen et al., 2022) investigated the impact of participant and NH characteristics on completing this intervention.This study showed that nursing profession (vs.physicians, physician assistants, nurse practitioners) significantly increased the likelihood of attendance by more than 5-fold, while early participation in the intervention and highest NH rating score increased overall adherence.
Another antibiotic stewardship intervention for nursing staff, physicians, healthcare assistants/helpers and caregivers (Hartman et al., 2023) showed that the intervention effect (on the number of prescriptions for suspected UTIs) was stronger in PwD compared to people without dementia.
In a third identified intervention to reduce antibiotic use, nurses and other healthcare assistants/helpers, (Kousgaard et al., 2022) perceived challenges in using the tool due to PwD´s communication difficulties and that external actors from the hospital sector and the municipality favored urinalysis, which was not aligned with the intervention approach (usually solved after explaining the rationale behind intervention approach).
A study investigating a decision-making algorithm to assist the management of suspected infections (Hughes et al., 2020) showed that nurses and other healthcare assistants face challenges to recognize urgency, abdominal pain, and urinal blood in PwD.They were also concerned about missing all behavioral changes that indicate infection in PwD, they identified challenges in using the tool due to PwD´s communication difficulties and perceived that the algorithm should differentiate behavioral changes for PwD and those without dementia.A consensus group from the same study displayed concerns about relying only on PwD's temperature, and they feared the algorithm´s applicability for PwD and its use among non-nursing staff.
The Illness Warning Instrument (Boockvar et al., 2000) aimed to improve communication between nursing assistants and other healthcare staff about acute illness signs, including infections.Instrument non-completion was attributed to lack of time, and most nursing assistants found the tool useful to facilitate communication.The tool effectively predicted illness likelihood within a week, showed high reproducibility, and validly assessed global status change.The last 2 identified interventions were a tool to detect early-stage infections (EDIS) used by nursing assistants (Tingström et al., 2015) and a machine learning algorithm that indicates UTIs in PwD's home (Enshaeifar et al., 2019) (further detailed in Table 2).

Discussion
We identified 22 studies that investigated decision-making processes (14 studies) and/or interventions targeting such processes (8 studies) for the detection and management of suspected infections in PwD.Almost all studies were conducted in NHs or similar care settings, and were mostly based on data from the US, Sweden, and the UK.Around a third of all studies drew upon three central source populations (SPREAD, EDIS, and TRAIN-AD).
The decision-making processes identified all fell under the themes of recognition of infections based on observation of early signs and symptoms, actions when suspecting infections, and proxy/family involvement.Interventions targeting decision-making processes included antimicrobial stewardship and other decision-making support tools.Determinants included healthcare staff perceptions, and different system, patient, family/proxy, and healthcare staff-related factors.Almost all the identified actors actively involved in the decision-making processes/interventions were healthcare staff.Only one study included PwD, and family/proxies were included in five studies in total.

Decision-making processes
Healthcare staff were primary actors in recognizing infection signs and symptoms, (Kovach et al., 2010;Allemann and Sund-Levander, 2015;Boockvar et al., 2000;Tingström et al., 2010;Chan et al., 2021;Harbin et al., 2022;Hartman et al., 2022;Jones et al., 2020;Sund-Levander and Tingström, 2013;Juthani-Mehta et al., 2009) and they reported several related challenges which are known and emphasize the complexity of infection detection and management in PwD and the need for interventions.However, we found that such challenges are potentially unmet in the developed interventions which we identified (discussed further below).
In the included studies in our review, several indicators for infection were highlighted such as fever and mental and behavioral change, but there were disagreements and challenges related to these (e.g., differing opinions on what constitutes fever, difficulty in recognizing behavior changes).Such findings are consistent with a previous study (Mayne et al., 2019) that showed lack of clarity in infection indicators for the general population of older adults, including confusion.All this points to the lack of unified understanding of infection indicators among healthcare staff and further raises questions on whether it is possible and beneficial to unify such indicators, particularly in the presence of the reported interpersonal difference among PwD (Tingström et al., 2010).
Moreover, the scarcity of identified studies on family/proxy and PwD involvement in decision-making warrants consideration and may hinder strategies that aim at improving infection detection and management for PwD.Decision-making can be burdensome for caregivers of PwD (Givens et al., 2012) who often deal with complex decisions, but understanding their experiences and needs is key to providing them with the best support and guidance (Moermans et al., 2022).Furthermore, the involvement of PwD in decision-making is continuously recognized as essential for personhood, person-centered care, and for their autonomy which subsequently maintains their quality of life and well-being (Lanzi et al., 2017;Taylor et al., 2023).
The scarcity of studies on community-dwelling older adults is also worrisome.Decision-making processes will normally be different in people living at home, with milder dementia, and younger.In our previous observational nationwide study, we have shown that infectionrelated hospitalization rates are even higher among younger PwD (Janbek et al., 2021a).Thus, exploration of decision-making process that can inform interventions in this group is of high relevance.For community-dwelling older adults, the daily surrounding environment does not necessarily include nursing healthcare staff (whom majority of our identified studies investigated) and therefore key factors determining the processes involved in detecting infections remain unknown and are emphasized by the lack of identified studies in our review.

Interventions
Only three of the included intervention studies investigated healthcare staff attitudes and experiences with the intervention, (Boockvar et al., 2000;Hughes et al., 2020;Kousgaard et al., 2022) and none of the identified interventions explicitly addressed the challenges PwD (including communication difficulties) and healthcare staff (including noticing behavioral changes) face.This highlights the need for more practice-based research on interventions, that ensures normalization and sustainability of the developed intervention in practice.Further, all but one (Boockvar et al., 2000) interventions included were specifically developed for infections.However, we question if infections should be addressed as a separate comorbidity with special detection tools/checklists, or together with other illnesses.A study on checklist fatigue in healthcare recommends reducing their number and designing them more carefully (Grigg, 2015).

Highlighted need to focus on dementia
In the public health realm of elderly care, there is a highlighted need to focus specifically on dementia.In our review, we found evidence from several studies emphasizing this need.Studies have highlighted the importance of knowing the PwD, (Sund-Levander and Tingström, 2013) the importance of considering PwD's voice in infection treatment decision-making, (Harbin et al., 2022) and the importance of distinguishing changes in behavior for PwD and without dementia to recognize infections and ensuring that developed detection tools differentiate this as well (Hughes et al., 2020).Also, evidence revealed reasons for the dementia-specific complexities involved in recognizing infections, including PwD exhibiting atypical signs and symptoms of infection, (Löppönen et al., 2004;Limpawattana et al., 2016) and having difficulties communicating their symptoms (Walker et al., 2000;Miller, 2001).Many studies have added to the evidence of dementia-specific infection signs and symptoms by presenting quotes as "…especially in dementia…" (Chan et al., 2021;Harbin et al., 2022;Hartman et al., 2022;Hughes et al., 2020;Jones et al., 2020).This underscores the need for infection detection and management approach to differentiate PwD from the general population of older adults.

Implications, highlighted gaps, and future research
The evidence synthesized in this review may serve as guidance for clinicians, public health practitioners, and policy makers to understand the complexity of dementia care and address needs of dementia actors expressed in our findings.One example is suggested by the highlighted emphasis on dementia and on the importance of knowing the PwD for a long period to help recognize indicators of an infection.This calls for the need of supporting nursing personnel in NHs and addressing their needs, including addressing a known pressing issue of rapid staff turnover (Shen et al., 2023;Castle and Engberg, 2005).
We highlight main gaps in evidence.First, the lack of studies outside of the NH and advanced dementia.Second, a big proportion of the included studies was conducted in the same three countries/ same source population, limiting the knowledge on practices between different healthcare models.Third, the lack of studies on PwD's involvement, and the scarcity of studies on family proxies.Following our current work, we urge researchers to cover such gaps in knowledge to continue our efforts in understanding what goes on in infection detection and management in PwD.

Strengths and limitations
Our review had several strengths.An iterative search approach supported a comprehensive overview of available practices involved in PwD care.Moreover, the iterative process of final article inclusion enabled us to include studies which did not include PwD as majority population (snippets) but from which we extracted relevant information on PwD that answered our research questions.Inductive thematic clustering of results helped us identify themes and synthesize evidence from all included studies.We gathered evidence drawn from real-world implementations to bring more effective and informed healthcare strategies (Glasgow et al., 2012).Therefore, we excluded studies not providing an evaluation of intervention effectiveness or the perceptions of healthcare staff implemented in practice.Moreover, we managed to explore several different interventions, outside of antimicrobial prescribing practices (Tingström et al., 2015;Enshaeifar et al., 2019;Boockvar et al., 2000;Hughes et al., 2020).Finally, we did not include grey literature to keep the quality standard of evidence-based practices among the included studies.
One limitation of our review was that due to the different study designs and differing exposures and outcomes, identified interventions could not be compared by their effectiveness in infection detection and management practices or methodological quality.Another limitation was that due to the large number of articles included with the same source population, we have an over-representation of certain practices and dementia actors.
In conclusion, this review highlights significant gaps in infection detection and management in PwD, particularly in recognizing symptoms and involving PwD and their family caregivers in decision-making.
Challenges include healthcare staff's need for familiarity with PwD to detect signs of infections and a lack of standardized indicators.The minimal research focus on community-dwelling PwD and the need for dementia-specific approaches were also emphasized.Our findings inform and encourage future research which should prioritize inclusive and practical interventions that address these gaps and improve healthcare strategies for PwD.

Declaration of Competing Interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Mihaela Ivosevic reports financial support was provided by Novo Nordisk Foundation.If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.The scope of the study and the research questions.The figure is a graphical presentation of the scope of this study and the research questions: a) Scope: the three boxes represent stages of infection prevention, detection, and treatment, with this study focusing on the middle box (detection and management).The scope thus includes all decision-making processes and interventions targeting these processes for suspected infections.b) Research questions: This study investigated the following: What are the decision-making processes in the detection and management of suspected infections?What interventions that target these processes exist?Which actors are included?What are the various determinants considered?

Fig. 4 .
Fig. 4. Mapping of the interventions targeting decision-making processes, determinants, settings, and actors.PwD-People with Dementia.TRAIN-AD -The Trial to Reduce Antimicrobial Use in Nursing Home Residents with Alzheimer Disease and Other Dementias.
Looks at physical signs (including physiological changes) and behavioral changes Outcome: Supervised model (bathroom usage, temperature) had 4 % Actors: Healthcare staff (continued on next page)

Table 1
Characteristics of included studies.
(Kovach et al., 2010).US Describe problems developing in NH residents and the relationship of patient/nurse factors to time from symptoms to problem identification, RCT analysis additionally using medical records NH 65 NH residents/PwD b , 15 nurses Advanced, any (Helton et al., 2011).US, Netherlands Study whether physician presence in the NH related to clinical decision making, cross-sectional study using online surveys NH 62 Physicians NA, pneumonia (Sund-Levander and Tingström, 2013).c Sweden Describe nursing assistant's experiences of the clinical decision-making process when they suspect an infection, qualitative study using focus group interviews (continued on next page)

Table 1
(continued ) (Juthani-Mehta et al., 2009)f included studies and their characteristics.Abbreviations: EDIS -Early Detection Scale of Infection, LRTI -Lower Respiratory Tract Infection, NA -Not Available, NH -Nursing Home, PwD -People with Dementia, RCT -Randomized Controlled Trial, TRAIN-AD -The Trial to Reduce Antimicrobial Use in Nursing Home Residents with Alzheimer Disease and Other Dementias, UK -United Kingdom, US -United States, UTI -Urinary Tract Infection a regarding snippets, this table consist of the entire study population for the articles included, but the synthesis only mentions actors related to the presented snippets b in the referenced studies, while the primary population was NH residents, a majority had dementia.To accurately represent this demographic, they are referred to as NH residents/PwD.cpart of the longitudinal project on early signs and symptoms and biochemical markers in NH residents with suspected infections d part of the Study of Pathogen Resistance and Exposure to Antimicrobials in Dementia (SPREAD) e in addition to acute illness f part of the trial to reduce antimicrobial use in NH residents with Alzheimer disease and other dementias (TRAIN-AD) (UTI) suspicion,(Juthani-Mehta et al., 2009)nurses and physicians listed a change in character of urine and fever/chills as the most common reasons (of physical signs; in 16 % and 13 % of 399 episodes, respectively).3.2.1.1.2.Observation: Behavioral and mental status changes.Nursing assistants

of infections based on observation of early signs and symptoms
Observation started 72 % of episodes (nothing happens in the rest) -Contact to physician ended 73 % of episodes and follow up in the others -Nothing happens was coded in 45 % of the episodes, and was more common when the nursing assistant was the initiator Nursing assistants had initiated episodes more often than the nurses -observation of physical signs and symptoms, and observation of behavioral and mental status changes Proxy/family involvement -PwD/family prompt to suspect infections Actors: Healthcare staff (physicians, nurses), proxy/family, PwD (continued on next page) M. Ivosevic et al.Ageing Research Reviews 101 (2024) 102520

Table 2
(continued ) Table 2 provides an overview of the summarized study results.Abbreviations: AOR -Adjusted Odds Ratio, CI -Confidence Interval, DNH -Do Not Hospitalize, NH -Nursing Home, OR-Odds Ratio, PwD -People with Dementia, RTI -Respiratory Tract Infection, UTI -Urinary Tract Infection