The scientific evidence for a potential link between confusion and urinary tract infection in the elderly is still confusing - a systematic literature review

The review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [13]. Inclusion criteria and methods were specified in advance and documented in the review protocol. The initial protocol was published with PROSPERO International prospective register of systematic reviews, registration number CRD42015025804.

This review included quantitative studies which met the following inclusion criteria:

The following exclusion criteria were applied:

The primary outcomes of interest were confusion, UTI and bacteriuria. Due to the large variety of terminology that surrounds the symptom of confusion, definitions used for confusion in this review encompassed: confusion, acute confusional state, delirium, altered mental status, altered mental state. This broad definition was used so as to capture all studies which may have assessed confusion but used alternative terminology. No absolute lower age limit was set, as it was anticipated that these would have an overall negligible effect on the data. This was so as to not exclude studies which may present data representative of an elderly population but which included a small minority of non-elderly participants. Studies that referred to specific subpopulations, or exclusively reported catheter associated UTI were also excluded so as to explore the association between confusion and UTI or bacteriuria in a general elderly population.

Three databases were accessed to identify studies eligible for this review: PubMed, Scopus and PsycINFO (via ProQuest). The search terms were related to three key topics: confusion, UTI and bacteriuria (Table 1) with adaptations for Scopus and PsycINFO. No restrictions on date or language were applied and studies published up to August 2015 were included. Once the final list of full text articles was determined, the references and citation history of the included studies were screened for other potential studies eligible for the review. The full texts of any new studies deemed possibly eligible for inclusion, were then retrieved and assessed.

After each database had been searched, the search results were collated and duplicates removed. Titles and, where available, abstracts retrieved were assessed for eligibility against the described inclusion criteria. Studies that clearly did not meet the review’s criteria were excluded. The full texts of potentially eligible studies and those that after title and abstract screening were not able to be definitively excluded were retrieved. The full text articles were then assessed for eligibility by the first reviewer (SM). Studies that could not be definitively excluded were discussed and resolved by consensus with another reviewer (RG). Studies excluded at this stage were recorded and their reason for exclusion is reported.

Data extraction was performed by one author (SM) using a standardised, pre-designed data extraction form, with the exception of data relevant to the quality assessment which was extracted by two review authors (SM and AB) independently. Any discrepancies identified were resolved by consensus, or through discussion with the third reviewer (RG). Data extracted from each eligible study included:

Two review authors (SM and AB) assessed the risk of bias of included studies independently, with any discrepancies being resolved by consensus, or through discussion with a third reviewer (RG), if necessary. The risk of bias was assessed using a modified version of the assessment checklist developed by Downs and Black [14]. Quality items that pertained to interventions and trial studies were removed as they were not deemed to be appropriate for the studies included in this review. An additional five quality items were added to the quality assessment to determine if studies described the criteria used for confusion, UTI and bacteriuria, and if their criteria for UTI and confusion were valid and reliable. Criteria for confusion were deemed valid and reliable if accepted criteria were utilised, including: the Confusion Assessment Method, the Organic Brain Syndrome Scale or the Diagnosis and Statistical Manual (DSM) criteria [15‐17]. Similarly, criteria for UTI were deemed valid and reliable if established criteria for UTI were utilised, including: the McGeer Criteria, the revised McGeer Criteria, the Loeb Criteria, or the Revised Loeb Criteria [1, 8, 18, 19]. The modified checklist finally consisted of 14 quality items, grouped into: reporting, internal validity, external validity and criteria (Table 2). The risk of bias for each quality item was reported as low risk of bias, high risk of bias, unclear risk of bias or not applicable.

Although the broad search strategy described was employed to enable the meta-analysis of data from included studies if deemed feasible, due to the heterogeneity of the data and the variety of definitions for confusion and UTI reported between the studies, meta-analysis was not conducted. Alternatively, a qualitative synthesis of the findings from the included studies was performed, structured around the quality of data, consistency of definitions and the evidence for the association between confusion and UTI.

Searches identified a total of 1907 records (Fig. 1). After duplicate records were removed, 1722 remained. These articles were then screened by title and abstract against the inclusion and exclusion criteria, leading to the exclusion of 1657 articles, as it appeared they clearly did not fulfil the eligibility criteria. Eleven potential records were excluded as their full texts were unable to be obtained. The full texts of the remaining 54 articles were closely examined. Thirty-five articles were excluded at this stage, with the most common reason being the study reported confusion and UTI/bacteriuria in their population, but not in relation to each other [20‐34]. Other common reasons included: the study did not report confusion [35‐39], the study did not report UTI/bacteriuria [2, 40‐43], the study only reported the association in a specific subpopulation [44‐46], or the study combined their measurement of confusion with other parameters [7, 47‐50]. Two studies were also excluded as UTI and confusion were not assessed concurrently [51] and reported UTI was combined with other parameters [52]. Three additional studies that met the inclusion criteria were identified by searching the references of relevant articles and searching for studies that cited these articles. No studies were deemed suitable for quantitative synthesis. A total of 22 articles met the inclusion criteria and were included in the systematic review.

The quality of the studies included in this review varied considerably (Fig. 2). This is partially due to inclusion of all studies that reported data on the rate of confusion in patients with suspected UTI or bacteriuria, or vice versa, even if it was not the main objective of the study. Reporting of main outcomes, description of patient characteristics and number of non-responders/patients lost to follow up were done well in most studies, with only one small study not clearly describing their main outcomes [53] and four studies not reporting the number of non-responders or patients lost to follow up [40, 54‐56]. In terms of internal validity, all applicable studies were deemed to have used appropriate statistical tests; however, half of the studies did not clearly describe other principle confounders in their comparison groups. The external validity, however, of all studies, was generally very high.

The quality of the criteria used to define UTI, bacteriuria and confusion varied considerably and was generally quite poor. No studies used criteria for UTI completely consistent with the revised Mcgeer or Loeb Criteria. Two studies employed a reproducible definition of UTI although neither employed published explicit criteria developed through expert consensus. [57, 58]. Many studies utilised discharge coding from patient databases which resulted in the reliability of their criteria being unable to be determined [59‐65]. Two studies were identified that used criteria that were clearly inappropriate [66, 67]. Three studies did not provide a definition for UTI, as they reported confusion in association with validated criteria for bacteriuria only [3, 68, 69]. Only one of these studies utilised an appropriate definition of bacteriuria and validated criteria for delirium [3]. Three of the studies which provided a definition for UTI also defined criteria for bacteriuria [56, 58, 67].

Almost all studies provided a definition of confusion criteria, but only eight studies used criteria that were valid and reliable (Table 3) [3, 54, 55, 60, 66, 70‐72]. Five studies used criteria for confusion which were clearly not valid or reliable [65, 67‐69, 73], and nine were unclear in their criteria used (Table 4) [53, 56‐59, 61‐64].

There were four large retrospective cross-sectional studies, and among the remaining studies the number of patients in each study varied considerably from small community samples of 9 to larger hospital samples of 710 (Tables 3 and 4). The majority of the studies identified were cross-sectional in design. Approximately half of the studies had an entirely elderly population ≥ 65 years (n = 10), with the other half of studies having populations deemed to be representative of an elderly population by median or mean age ≥ 65 years (n = 10). In the two remaining studies, one conducted in a nursing home and the other in a psychogeriatric unit, the demographics of the patient sample were not provided. They were believed to be representative of an elderly population by their care setting. The proportion of participants with urinary catheters was unclear in the majority of included studies (n = 14). In the remaining studies, urinary catheter rates were high, 37–51% (n = 3), low 1.8–5.5% (n = 2) and none (n = 3). The majority of the studies were conducted in a hospital setting (n = 14), followed by nursing homes (n = 6) and community settings (n = 2). Interestingly, only two of the included studies had the explicit aim of exploring the association between confusion and UTI; however, ten studies did partially explore this association.

Among the included studies, the rate of confusion in patients with suspected UTI was most commonly reported (n = 13) followed by the rate of suspected UTI in patients with confusion (n = 10). Some studies reported the rate of confusion in patients with bacteriuria (n = 4) and one study reported the rate of bacteriuria in patients with confusion. The majority of studies reported confusion as delirium (n = 15), followed by a few reporting confusion (n = 3), altered mental state (n = 2), and mental status changes (n = 1), with one study reporting both delirium and altered mental status [57]. Twelve studies analysed the correlation between suspected UTI or bacteriuria and confusion (Tables 3 and 4).

No study used validated definitions of both confusion and UTI, so this association could not be reliably established. Only one study by Juthani-Metha et al. had an acceptable definition of confusion and an acceptable definition of bacteriuria. They found an association between bacteriuria and confusion with the relative risk being 1.4 (95% CI 1.0–1.7, p = 0.034) [3].

Following this review, it is evident that all of the studies which have explored the association between suspected UTI and confusion are methodologically flawed, due to poor case definition for UTI or confusion, or inadequate control of confounding factors introducing significant bias. Subsequently, no accurate conclusions about the association between UTI and confusion can be drawn. One study of acceptable quality shows an association between confusion and bacteriuria. However, this sample of patients in whom they tested bacteriuria and pyuria were patients already suspected of having a UTI, introducing a bias into their calculation [3]. In summary, none of the 22 publications had sufficient methodological quality to enable valid conclusions.

Frail residents are more likely to have bacteriuria [74]. Frailty also predisposes for cognitive decline [75, 76]. Hence, there might be an indirect link between confusion and bacteriuria, easily misinterpreted as UTI causing confusion. This might explain some of the trends found in the existing literature.

Studies including hospitalised patients are likely to also include patients with pyelonephritis, a condition likely to result in confusion in a fragile elderly person. However, the typical nursing home situation usually involves the suspicion of confusion caused by a lower UTI (acute cystitis) in an afebrile patient.

The primary aim of this review was not to evaluate the association between pyelonephritis and confusion. The primary question was if lower UTI with no fever in residents without a urinary catheter, with or without localised symptoms such as acute dysuria, urgency or frequency, is associated with confusion. This review concludes that current evidence does not provide a clear answer to this question.

The strengths of this review are mainly due to its methodological quality; that it utilised a broad search strategy, with no limits to age or date applied. This allowed for studies that were representative of an elderly population and without the explicit aim of reporting the relationship between confusion and UTI to be identified. Another strength of this review was the registration of a protocol with pre-specified objectives and methods. The use of a second reviewer independently assessing the quality of selected studies also increases the quality of the review. Limitations included limiting articles to English and being unable to assess the eligibility of the unobtainable full-texts. This review also did not attempt to include studies from the unpublished literature, introducing possible publication bias.