Protocol and registration

This systematic review and meta-analysis examined the prevalence of thirst, assessment methods, risk factors, and symptom relief methods in critically ill patients. The study methodology followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Statement [9], and a systematic review was conducted. The protocol was registered in PROSPERO (ID: CRD42023428619) on June 6, 2023 (S1 File). The systematic review adhered to the PRISMA 2020 checklist (S2 File).

Literature search strategy

We searched for articles in three electronic databases: MEDLINE via PubMed, the Cochrane Library (CENTRAL), and CINAHL. We also conducted a manual search using Google Scholar. All English publications published until June 6, 2024 were searched without any restrictions on countries. The reference lists of all selected articles were independently screened to identify additional studies that were excluded from the initial search. The search formulae for PubMed, CENTRAL, and CINAHL are presented in Supplementary S3 File.

Study screening and selection

Two of the five reviewers (TF, GA, SO, SU, and HS) independently screened the titles and abstracts of all studies to identify potentially relevant studies. Subsequently, the full texts were independently reviewed according to a standardized protocol. Any disagreements between the two reviewers were resolved through discussion, and if necessary, a third reviewer (AO) was brought in for arbitration.

Eligibility criteria

To conduct a comprehensive and exploratory review of thirst in critically ill patients, we used the “population–concept–context” framework [10] recommended by the Joanna Briggs Institute as a guide for developing clear and meaningful objectives and eligibility criteria; The study population included critically ill adult patients expected to stay in the ICU for more than 24 hours, the concept was any treatment for thirst or dry mouth, and the context was critically ill patients in ICUs. The study design included cohort, cross-sectional, and prospective studies, including randomized controlled trials (RCTs) and non-RCTs (NRCTs) with control groups. Only studies reporting the prevalence of thirst, risk factors, and treatment methods for thirst in critically ill patients, respectively; that did not include preclinical results; were in English; and that included human participants were included.

Data extraction and quality assessment

The prevalence of thirst was recorded as the primary outcome of each study. Secondary evaluations were collected from the literature and included assessment methods and risk factors for thirst, and symptom relief methods. Data extraction was performed by TF. Data were extracted using Microsoft Excel. Each article was summarized under the following data headings in an Excel spreadsheet: First author; Year of publication; Title of Article; Study characteristics; Sample size; Control group; Prevalence; Demographic factors (including average age, severity of illness, etc.); Symptom relief methods; Risk factors; Assessment methods; and Definition of thirst.

The risk of bias in the target literature was assessed using the Cochrane Risk of Bias 2 tool (RoB2) for RCTs [11], and the study design’s quality and degree of potential bias were assessed according to the RoB2 domains. Two reviewers (TF and AO) independently evaluated the risk of bias using the RoB2. Disagreements between the two reviewers were discussed and if unresolved, a third reviewer (HS) was brought in for arbitration. The Newcastle-Ottawa Scale was used for observational studies, cross-sectional studies, and NRCTs. For RCTs, we assessed each of the following six domains: (1) selection bias in the process of randomization assignment, (2) selection bias in the process of concealment of assignment, (3) implementation bias in the process of blinding to participants and interventionists, (4) measurement bias in the process of blinding to outcome assessors and analysts, (5) missing bias due to participant attrition, and (6) reporting bias in the reporting of outcomes were assessed. For observational studies, cross-sectional studies, and NRCTs, we assessed each of the following three domains: (1) representativeness, exposure assessment, etc. (four items), (2) confounding assessment (two items), and (3) outcomes (three items). For observational studies, cross-sectional studies, and NRCTs, the risk of bias was assessed according to each of the following three domains: (1) representativeness, exposure assessment, etc. (four items), (2) confounding assessment (two items), and (3) outcomes (three items). The risk of bias for the Newcastle-Ottawa Scale was assessed by summing the scores of the nine criteria to evaluate the overall quality of each study; if the two raters had different opinions, a decision was made through an inter-rater discussion.

Statistical analysis

Analyses were conducted via Stata/BE (version 18.0; Stata Corp., College Station, TX, USA). The point estimates from each study were combined using a random-effects meta-analysis model to obtain an overall estimate using the Der Simonian–Laird method. The Freeman–Tukey double-arcsine transformation was chosen as an effect size measure. The estimate was considered significant when the Z test p-value was < .05. We assessed heterogeneity with Cochrane’s Q test and tau-squared (T²) and measured the inconsistency (the percentage of total variation across studies due to heterogeneity) of effects across interventions using the I² statistic. Publication bias was assessed using a funnel plot and Egger’s test, with significance set at P < .05. In addition, the prevalence of each definition of thirst sensation was extracted by subgroup analysis. A meta-regression analysis to determine the contribution of specific cofactors (mean age, sex ratio, mechanical ventilator use, and length of ICU stay) to heterogeneity was conducted.

Literature search results

The literature search results are summarized in Fig 1. A total of 1,267 references were retrieved from the three English databases. A total of 1,093 references were targeted for primary screening after excluding duplicates. Subsequently, 100 studies were selected and evaluated for full text as secondary screening. Fifteen eligible studies met the inclusion criteria.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Preferred reporting items for systematic reviews and meta-analyses statement flowchart.

https://doi.org/10.1371/journal.pone.0315500.g001

Selected literature characteristics

After the screening, 15 studies were selected for eligibility. Regarding the study design, nine studies were observational [4,12–19], two studies were cross-sectional [6,20], and four studies were RCTs [7,8,21,22]. Six studies were from the United States [4,7,12,13,16,20], two each from France [14,22], Japan [6,15], and China [8,18], and one each from Italy [17], South Korea [21], and Norway [19]. Two references were published in the 2000s [12,13], seven in the 2010s [4,6,7,14,16,20,21], and six in the 2020s [8,15,17–19,22].

Regarding patient characteristics of the selected references, five studies included only patients on ventilators [13,16,17,19,21] and four studies had a mean patient age of 65 years or older [6,12,14,15]. Furthermore, eight references were mixed ICUs [6,7,13,15,18–20,22], two were internal medicine ICUs [12,21], three were general ICUs [8,14,17], one was a surgical ICU [16], and another was an ICU with no type description [4]. Some internal medicine ICUs were included in the literature, thus limiting their scope to patients with cancer (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study characteristics.

https://doi.org/10.1371/journal.pone.0315500.t001

Risk of bias assessment

Risk of bias was assessed using the RoB2 and the Newcastle-Ottawa Scale and are summarized in Table 2. Finally, two studies were judged to be high risk, ten to be with some concerns, and three to be low risk.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Quality rating of studies reporting.

https://doi.org/10.1371/journal.pone.0315500.t002

Two reviewers independently assessed the risk of bias for each included study. When the Cochrane Collaboration Risk of Bias Tool was used, the overall risk of bias was determined by following the tool’s standardized instructions: studies assessed to have “low” risk of bias in all tool domains were considered to have “low” overall risk, studies assessed to have “some” risk of bias in at least one tool domain were considered to have “some” overall risk, and studies assessed to have “high” risk of bias in at least one domain or assessed to have “some” risk of bias in two or more domains in a way that “substantially lowers confidence in the result” were considered to have “high” overall risk. When the Newcastle-Ottawa Scale was used, the overall risk of bias was determined by the average score of the two reviewers: greater than or equal to seven was considered as having “low” overall risk, greater than or equal to four but less than seven was considered as having “some” overall risk, and less than four was considered having “high” overall risk.

Primary outcomes

The prevalence of thirst is summarized in Fig 2. Excluded studies in this analysis are summarized in S1 and S2 Tables. Eight studies [4,6,12,15,17–20] reported the prevalence of thirst drawing on data from 2,204 patients, resulting in a prevalence estimate of 0.70 (95% CI: 0.59–0.79), with high statistical heterogeneity (T² =  0.09, I² =  95.45%, P < .001). When stratified by the definition of thirst, there were four references with a numerical rating scale (NRS) score ≥  1 or with other scales to determine the presence of thirst. The prevalence estimate was 0.71 (95% CI: 0.67–0.76), with low statistical heterogeneity (T² =  0.00, I² =  37.96%, P = .18). Three references with an NRS score ≥  3 had a prevalence estimate of 0.77 (95% CI: 0.66–0.86), with high statistical heterogeneity (T² =  0.03, I² =  88.45%, P < .001). One had an NRS score ≥  8 and a prevalence estimate of 0.41 (95% CI: 0.36–0.46); heterogeneity could not be assessed. For excluded study is summarized in S4 File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Forest plot for prevalence of thirst sensation.

https://doi.org/10.1371/journal.pone.0315500.g002

The risk factors for thirst are summarized in Table 3. Risk factors for thirst were divided into patient and treatment factors. Four patient factors were identified: serum sodium concentration, hyperglycemia, illness severity, and xerostomia. Six treatment factors were identified: nil per os, surgery, diuretics, humidified Venturi mask, serotonin reuptake inhibitors, and high doses of opioids in morphine equivalents (>50 mg).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Risk factors for thirst in critically ill patients.

https://doi.org/10.1371/journal.pone.0315500.t003

Secondary outcomes

The methods used to evaluate thirst are summarized in Table 4. Thirteen studies reported methods of assessing thirst, with NRS being the most frequently evaluated method by nine references [6,7,8,13,15,17,18,23]. The visual analog scale (VAS) was reported by one study [14], the Edmonton Symptom Assessment Scale by two studies [4,12], the Vidatak EZ picture board by one study [16], Patient Symptom Survey symptom checklist [19] and the Tobii Dynavox I-15 eye tracking device by one study [16].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Assessment method and definition of thirst sensation included in the systematic review.

https://doi.org/10.1371/journal.pone.0315500.t004

Symptom relief methods for thirst are summarized in Table 5. Five articles reported methods for alleviating the symptoms of thirst. Doi et al. [15] examined changes in the NRS for thirst, lip moistness, and the Modified Revised Oral Assessment Guide after providing oral care every four hours to the intervention group. The results showed a significant effect on the NRS of thirst immediately after oral care (-1 [-3 to 0], P < .01) to 1 h later (0 [-1.25 to 1], P =  0.04). Jang et al. [21] performed tooth brushing, swabbing with 0.1% chlorhexidine, and intermittent swabbing with cold water for the intervention group and conventional oral care using gauze and saline for the target group, and compared them with the Beck Oral Health Assessment Scale. The results showed significant effects on oral status (8.1 vs. 11.1, P < .001), dry mouth (sublingual: 2.0, 1.0, P < .001; dorsal tongue: 1.0, 0.0, P < .001), and pH (6.5, 5.7, P < .001). Puntillo et al. [7] compared the intensity of thirst and distress in an intervention group with 15 min of oral swabbing, water spray, and application of menthol moisturizer to patients’ lips and in a control group with 15 min of usual care. The results showed significant effects on both thirst intensity (-2.3 vs. -0.6) and distress (-1.8 vs. -0.4). In a RCT [8], the intervention group received hourly oral moisturization with vitamin C spray, mouthwash with peppermint water, and application of lip balm. Oral moisturizer with saline spray every hour, mouthwash with warm water, and lip moisturizer with warm water every 2 h were given to the control group, and NRS of thirst feeling and the objective oral mucosa scale were compared. The study results showed significant effects on both the NRS of thirst (-1.27 vs. -0.19, P < .001) and the objective oral mucosa scale (-0.36 vs. 0.1, P = .008). In addition, there were four references that used oral care bundles and interventions. On the other hand, Merliot-Gailhoustet et al. [22], who assessed pain, anxiety, thirst, dyspnea, and insomnia with virtual reality equipment and music therapy for the intervention group and usual care, such as television and radio for the target group, did not find a significant reduction in thirst.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Symptom relief methods of thirst in critically ill patients.

https://doi.org/10.1371/journal.pone.0315500.t005

Meta-regression analysis

Meta-regression analyses indicated no significant difference in the prevalence of thirst based on mean age, sex ratio, and mechanical ventilator use (S3 Table). We found a negative association between the overall prevalence of thirst and ICU length of stay.

Publication bias

The funnel plot is shown in S1 Fig. However, Egger’s test results did not uncover any significant indications of publication bias (z =  1.38, p = .17).

Prevalence of thirst in critically ill patients

The prevalence of thirst in critically ill patients was very high in this study. To our knowledge, this is the first systematic review to focus on the prevalence of thirst in critically ill patients. Konstam et al. [23] reported a 16% prevalence of thirst in medically treated patients with stable heart failure. Lee et al. [24] also reported a prevalence of moderate to severe thirst of 55.8% in postoperative patients. Compared to these studies, the prevalence of thirst in critically ill patients was higher.

Mechanisms can be considered regarding risk factors for thirst sensation in critically ill patients. The reasons for thirst in critically ill patients include an environment in which they cannot drink freely, such as during fasting, as well as a deficiency of intracellular fluid due to elevated plasma osmolality, extracellular fluid due to decreased blood volume, and mouth dryness [25–28]. An increase in osmolality is believed to cause the release of antidiuretic hormones from the posterior pituitary gland by stimulation of osmotic receptors in the anteroventral wall of the third ventricle of the hypothalamus, which sends stimuli to the anterior cingulate cortex and cerebellum, causing a feeling of thirst. ICU patients are prone to hyperglycemia [29], hypernatremia has been reported in approximately 26% of ICU patients [30], and diuretics are associated with increased osmolality. As for the decreased blood volume, stimulation of the arterial and baroreceptors of the heart and kidneys, resulting in atrial natriuretic peptide inhibition and sodium and water reabsorption, may send stimuli to the anterior cingulate cortex and cerebellum, causing a sense of thirst. Generally, ICU patients are prone to thirst sensations due to fluid balance changes, such as bleeding, vomiting, diarrhea, sweating, and diuresis. When the mouth is opened by oral tracheal intubation, xerostomia may cause the secretion of antidiuretic hormone from peripheral osmotic receptors in the oropharynx, which sends stimuli to the anterior cingulate cortex and cerebellum, causing a sensation of thirst. Regarding opioid effects, which are associated with the body’s water regulation system and may affect thirst perception [31,32], selective serotonin reuptake inhibitor suppress electrical activity in the subfornical organ at the entrance to the third ventricle; therefore, these drugs may act on this forebrain structure to promote sodium intake and thirst [33].

Regarding humidified Venturi masks, there is the impression that heating and humidifying is a positive effect, however, these masks are a risk factor for thirst sensation. Although fasting, eating, drinking, and other factors may have an influence in the background, heating and humidifying alone is unlikely to lead to a reduction in thirst. Based on the above findings, it is likely that the difference in prevalence is due to more factors that cause thirst in severely ill patients. Therefore, it is necessary to determine the prevalence of thirst at an early stage using appropriate assessment methods.

The assessment methods of thirst in critically ill patients

Based on the study results, the cut-off for the presence or absence of thirst using the NRS varies from NRS 1 or higher, NRS 3 or higher, to NRS 8 or higher, and the definition of thirst is not well defined. Owing to the subjective nature of the symptoms, it is difficult to define a clear cut-off. In any case, the optimal assessment of symptoms should include intensity, frequency, and distress [34], and should quantify more than just the presence or absence of thirst. Furthermore, in a systematic review examining strategies in the management of postoperative thirst [35,36], the NRS or VAS was the most common method of assessment. Similar to pain, the NRS or VAS method for evaluating thirst, a subjective symptom, is considered the gold standard. Recent studies have attempted to quantify the distress caused by dry mouth in detail using the Thirst Discomfort Scale, a 12-item five-point scale [37]. Thus, it is important to appropriately assess thirst for early detection, prevention, and symptom relief in critically ill patients.

The symptom relief methods of thirst in critically ill patients

The mechanism of thirst is thought to stem from a deficiency of extracellular and intracellular components and from stimuli sent from the oropharyngeal region to peripheral osmotic receptors. Studies examining the mechanism of thirst have reported [27] that cold stimulation of the oropharyngeal region is more useful than correction of dehydration in improving thirst. The results indicated that methods to alleviate thirst included oral moisturizing and ice water, citric acid, and ascorbic acid for pH adjustment, use of peppermint and menthol to provide a cool sensation, and lip moisturizing. Cold water ingestion may be more effective than room temperature water in relieving thirst as a stimulus to the oropharyngeal region [38,39], and ice water is representative of direct cold stimulation of the oropharynx. Other interventions, like citric acid and ascorbic acid, can promote salivation and alleviate thirst [40]. It has been confirmed that peppermint and menthol act directly on the oral mucosa and provide long-lasting cold stimulation, which promotes saliva secretion, retaining moisture in the mucous membrane, and alleviating the feeling of thirst [28,40]. Although the intervention methods are different, this intervention can achieve the effects of both cold stimulation and saliva secretion. Regarding lip moisturization, frequent glycerin application can maintain mucosal homeostasis [41]. Therefore, these mechanisms suggest that the symptom relief methods qualitatively integrated into this study were effective for thirst.

Furthermore, symptomatic relief from thirst requires a bundle of multiple interventions rather than a single one. In their study of symptom relief, Puntillo et al. [7] compared the intensity of thirst and distress in an intervention group with 15 min of oral swabbing, water spray, and application of menthol moisturizer to patients’ lips and in a control group with 15 min of usual care. They reported a 2.3 point reduction in thirst intensity and a 1.8 point reduction in thirst distress. In Zhang et al.’s study [8], the intervention group received hourly oral moisturization with vitamin C spray, mouthwash with peppermint water, and application of lip balm. Oral moisturizer with saline spray every hour, mouthwash with warm water, and lip moisturizer with warm water every 2 h were given to the control group. They reported a 1.27 point decrease in the NRS score for dry mouth. For patients experiencing pain, a decrease of approximately two points in the NRS score is considered clinically important [42]. Therefore, Puntillo et al. may highlight an effective intervention because it reduced the intensity of thirst by more than two points. However, a comparison of the post-care NRS scores between the intervention and control groups revealed a decrease of only 1.1 points. Thus, differences in intervention bundles, intervention procedures, timing of evaluation, inclusion and exclusion criteria, and other factors make it difficult to directly compare results across studies; however, no single bundled intervention would be able to accomplish much if it were less effective.

Limitations and implications

A strength of this study is that it is the first systematic review and meta-analysis to validate thirst in critically ill patients, allowing us to estimate its prevalence. However, the results showed high heterogeneity in the prevalence of thirst among critically ill patients. We estimated high heterogeneity based on the definition of thirst, with the group defined as NRS 3 or higher showing the greatest variability. In addition, meta-regression analyses results indicated that only length of ICU stay showed significant differences. However, further research is needed to fully explain the heterogeneity. Possible reasons for this include differences in patient populations, assessment methods, timing, inclusion and exclusion criteria, and other factors. Therefore, caution should be exercised when interpreting the integrated results regarding the prevalence of thirst in critically ill patients. As a challenge for future research, new surveys and studies on this subject are needed to obtain a more comprehensive overview as relevant studies are scarce. Second, since the study results showed high heterogeneity in the prevalence of thirst in critically ill patients, discussions about the definition of thirst, sample size, and inclusion criteria should be considered. Third, the criteria for the presence of thirst are defined differently by different studies, and it is unclear which values are clinically meaningful. Therefore, basic research is needed to unify thirst definitions in the future.