Which Frail Older People Are Dehydrated? The UK DRIE Study

Abstract

Background:

Water-loss dehydration in older people is associated with increased mortality and disability. We aimed to assess the prevalence of dehydration in older people living in UK long-term care and associated cognitive, functional, and health characteristics.

Methods:

The Dehydration Recognition In our Elders (DRIE) cohort study included people aged 65 or older living in long-term care without heart or renal failure. In a cross-sectional baseline analysis, we assessed serum osmolality, previously suggested dehydration risk factors, general health, markers of continence, cognitive and functional health, nutrition status, and medications. Univariate linear regression was used to assess relationships between participant characteristics and serum osmolality, then associated characteristics entered into stepwise backwards multivariate linear regression.

Results:

DRIE included 188 residents (mean age 86 years, 66% women) of whom 20% were dehydrated (serum osmolality >300 mOsm/kg). Linear and logistic regression suggested that renal, cognitive, and diabetic status were consistently associated with serum osmolality and odds of dehydration, while potassium-sparing diuretics, sex, number of recent health contacts, and bladder incontinence were sometimes associated. Thirst was not associated with hydration status.

Conclusions:

DRIE found high prevalence of dehydration in older people living in UK long-term care, reinforcing the proposed association between cognitive and renal function and hydration. Dehydration is associated with increased mortality and disability in older people, but trials to assess effects of interventions to support healthy fluid intakes in older people living in residential care are needed to enable us to formally assess causal direction and any health benefits of increasing fluid intakes.

Water-loss dehydration results from insufficient fluid intake, is indicated by elevation of directly measured serum osmolality, and undermines the health of older people. Consistent evidence from appropriately adjusted prospective studies suggests that dehydration is associated with increased mortality in stroke patients and older people with and without diabetes, and doubled 4-year disability risk (1–4).

Water is crucial to every bodily action, and maintenance of hydration is essential to life. Membrane osmoreceptors monitor water balance (5,6). In healthy young humans, restricted fluid intake leads to raised serum osmolality, triggering thirst leading to increased fluid intake, and releasing vasopressin (anti-diuretic hormone) which reduces fluid loss via the kidneys, increasing urine concentration. This restores water balance. Older people are thought to be at greater risk of dehydration as thirst sensation and urinary concentrating ability frequently decline with age (7–9). Many older people use diuretics or laxatives which encourage fluid loss, and reduced muscle volume leads to a smaller fluid reserve (10,11). Oral fluid intake may fall in older people for a variety of reasons (12,13), including reduced enjoyment of drinks, physical limitations, unmet activities of daily living needs, and decisions aimed at controlling continence (6,14,15). Additionally, those with dementia may forget to drink as daily routines are lost and social contact diminishes. This suggests that cognitive impairment, poor thirst, and poor renal function are likely to increase dehydration risk, but there is little direct evidence of this.

Hydration status needs to be assessed in older people using serum or plasma osmolality, the core physiological indicator of dehydration, as other measures become less useful and specific with increasing age (16–18). Creatinine-based measures are unhelpful in people with limited renal function and are nonspecific (also responding to cardiac failure, sarcopenia etc.) (17,19). Weight change (a crucial measure of dehydration in children, infants, and athletes) can be misleading as dehydration may occur slowly in older people, not triggering rapid weight change thresholds (20), while weight can fluctuate in well-hydrated older adults (21). Signs and tests for dehydration, such as skin turgor, dry mouth, capillary refill time, and urinary measures, appear to be poor indicators of dehydration or at best lack any evidence base in older adults (22). Although clinicians and researchers rarely routinely measure serum osmolality directly (using freezing point depression), they indirectly recognize its import by estimating it using a variety of osmolarity equations (23,24).

To protect older people living in long-term (residential) care from dehydration and its consequences, and to trial interventions to prevent dehydration, we need to understand its prevalence and which individuals are most at risk. There is limited and contradictory evidence of factors associated with dehydration in frail older adults (25). Studies that assessed associations with dehydration or low fluid intake using more reliable indicators (serum osmolality, tonicity, International Statistical Classification of Diseases and Related Health Problems (ICD) codes, or fluid intake over ≥24 hours) in older people identified greater age, female sex, and non-Caucasian ethnicity as associated with greater risk in community-dwelling and hospitalized older people (26–30), though sex and age were not related to hydration status in a large case–control study of hospital admissions (31). In 18 hospital-based patients, dementia was associated with dehydration (32) and poor cognition with low fluid intake in 57 long-term geriatric ward residents (33). Functional limitations were associated with increased dehydration risk and low fluid intake in community-dwelling older adults (26), and residents of geriatric units (33), but functional limitations, needing help with drinking, speech impairment, and drooling, were associated with improved fluid intake in 99 care-home residents (29). Health factors, including diuretic use (31), obesity, diabetes, hypertension, and chronic disease, have been associated with dehydration (26), whereas urinary incontinence, reduced nutrient intake, and fewer drinking sessions have been associated with low fluid intake (29). As limited sets of factors potentially associated with dehydration have been assessed in individual studies, often with very few participants and without appropriate adjustment for confounding, we aimed to assess a wide range of cognitive, functional, and health-based potential risk factors for dehydration (assessed by serum osmolality) in older people living in long-term care.

Methods

Dehydration Recognition In our Elders (DRIE) study methods are fully described in Supplementary File A and summarized here, the protocol and published study paperwork (23). DRIE was approved by the UK National Research Ethics Service Committee London–East Research Ethics committee (11/LO/1997; 25 January 2012), and all study procedures were in accordance with the ethical standards of the Helsinki Declaration.

DRIE included people aged 65 years or older living in long-term care (residential, nursing, and specialist-dementia homes) with written informed consent or written consultee agreement in England. Where residents were unable to demonstrate capacity to provide informed consent but expressed desire to participate, we asked their consultee (people who knew the potential participant well, usually a spouse, son, daughter, or long-term friend) to tell us whether the resident would have participated in our study if they still had capacity, and if so to give their written agreement. The consultee completed an opinion form on behalf of the resident, and we only included participants who had signed their own written informed consent, or where we obtained signed consultee agreement. Residents could withdraw consent, without providing reasons, at any point—verbally or through their behavior (by appearing not to want to converse or to take part in the interview).

We aimed for a representative sample of care-home residents, while recognizing we were likely to include higher proportions of those more physically and cognitively able. If residents were eligible (not receiving palliative care and the care home had no record of any diagnosis of renal or cardiac failure), had not told care staff they did not want to participate, and carers felt the resident was well enough, we asked whether they might like to take part. Where interested we discussed the participant information sheet, assessed capacity, and took consent or requested consultee advice.

Data were collected from care notes, staff, and resident interviews. Nonfasting venous blood was collected using needle and syringe, transferred to collection tubes, and delivered to the Department of Laboratory Medicine, Norfolk and Norwich University Hospitals Trust within 4 hours. Serum osmolality (freezing point depression; Advance Instruments Model 2020, repeatability ±3 [SD ±1] mmol/kg in the 0–400 mmol region, CV0.56), serum urea, creatinine, sodium, potassium, glucose (Abbott Architect), and hemoglobin (Instrument Sysmex XN) were measured in all available samples. Estimated glomerular filtration rate (eGFR) was calculated (34). Hydration status was classified by serum osmolality: normally hydrated (275 to <295 mOsm/kg), impending dehydration (295–300 mOsm/kg), and current dehydration (>300 mOsm/kg) (17,19).

Interview questions included EuroQoL-5D-3L (www.euroqol.org/), Mini-Mental State Examination (MMSE) (35), and short questions on feelings, drinks, sleep, continence, and exercise. Participants were asked whether they currently felt thirsty, just before venepuncture. We physically assessed participant’s mouths, body temperature, hands, feet, axilla, and eyes. Weight, vision (Snellen test), blood pressure, and pulse after sitting for ≥10 minutes (Omron M3) and after 1 and 3 minutes of standing were measured. Care staff provided information on recent, current and chronic illnesses, health care contacts, medications, weight history, functional status (Barthel Index), risk factors for poor food and fluid intake, or increased fluid requirements.

Data Analysis

The DRIE population was described by hydration status. We assessed participants’ representativeness compared with all care-home residents (living in homes included in DRIE) by age, gender, and body mass index (BMI). Univariate linear regression (STATA 11) was used to assess relationships between participant characteristics (age, sex, suggested dehydration risks, general health, markers of continence, cognitive and functional health, nutrition status, and medications used), 67 factors in all, and serum osmolality. As a sensitivity analysis, to assess the stability of findings to different statistical methods, we also used univariate logistic regression (STATA 11) to assess relationships between these characteristics and odds of impending and current dehydration. Where several characteristics within a category (categories shown in Supplementary Table 1) were statistically significant to p < .10 all were entered into multivariate linear regression and the characteristic with the largest p value removed stepwise until all remaining factors were p < .10 (this p value was chosen to ensure that in this limited data set we did not lose potentially important factors too soon from the analyses). All remaining characteristics, from all categories, were entered into the full backwards multifactorial regression model, and the characteristic with the largest p value removed each time until all remaining factors were p < .05. The same process was followed for multivariate logistic regression for both impending and current dehydration.

There are two groups of people with raised serum osmolality, those with and without raised serum glucose. For those without raised glucose, fluid intake is increased to correct raised osmolality, but for those with raised glucose (>7.8 mmol/L or >140mg/dl) treatment is primarily through diabetic control. We assessed whether risk factors for raised serum osmolality and dehydration altered when we omitted participants with raised serum glucose.

The strong correlation between poorer cognitive function and osmolality encouraged us to consider (post hoc) whether lack of thirst may be a mediator. We hypothesized that in the absence of thirst, it is easier to forget to drink, so we explored the relationship between feeling thirsty (participants were asked whether they currently felt thirsty, just before venepuncture) and osmolality, assessed using univariate linear regression.

Results

Interviews took place in 56 care homes from April 2012 to August 2013. Homes included 1,816 residents of whom 1,077 were deemed ineligible by care-home managers (Figure 1). Of the 739 potentially eligible residents approached by researchers, 374 were not interested and 365 wanted to take part, of whom 256 provided informed consent or consultee agreement. We initiated research interviews with 232 individuals and obtained serum osmolality for 201 individuals. Laboratory errors led to rejection of three serum osmolality readings and seven participants who had low osmolality and three participants who had cardiac failure (unknown before interview), leaving 188 included participants.

Figure 1.

Flow diagram for the Dehydration Recognition In our Elders (DRIE) study.

Participant Characteristics

The mean age of DRIE participants was 86 years (range 65 to 105 years), and 66% were women (Table 1, where characteristics are also displayed by hydration status). Mean serum osmolality was 293 mOsm/kg and 52% were well hydrated, 28% had impending dehydration, and 20% had current dehydration (Figure 2). Mean MMSE score was 22 (of 30, range 0 to 30) and 22% had normal cognition (MMSE > 26), 45% had mild cognitive impairment (MMSE 20–26), 26% had moderate cognitive impairment (MMSE 10 to <19), and 3% had severe cognitive impairment (MMSE < 10). Mean functional status was 67 (Barthel Index, range 0 to 100), 19% had diabetes, and 16% were underweight (BMI < 20kg/m²). Renal function was limited in many participants, with mean eGFR of 63mL/minute/1.73 m² (SD 18.6, range 18 to 90mL/minute/1.73 m²)

Table 1.

Characteristics of DRIE Study Participants

Characteristic	All 188 DRIE Participants	Hydration Status
		Current Dehydration	Impending Dehydration	Hydrated
	Mean (SD), n	Mean (SD), n	Mean (SD), n	Mean (SD), n
Serum osmolality, mOsm/kg	293.4 (8.1), 188	304.2 (3.5), 38	296.9 (1.6), 52	287.3 (5.5), 98
Sex, % Female	66% (124 of 188)	58% (22 of 38)	58% (30 of 52)	73% (72 of 98)
Age, years	85.7 (7.8), 188	84.6 (8.1), 38	85.3 (8.2), 52	86.2 (7.6), 98
MMSE Score (of 30)	21.8 (5.7), 180	19.3 (6.2), 34	21.6 (5.2), 48	22.6 (5.7), 96
Barthel Index (of 100)	67.4 (26.1), 188	61.4 (25.7), 38	69.0 (25.2), 52	68.8 (26.5), 98
Diabetes, %	19% (36 of 188)	37% (14 of 38)	23% (12 of 52)	10% (10 of 98)
Serum glucose, mmol/L	7.0 (3.2), 164	9.3 (4.9), 33	6.9 (2.6), 44	6.2 (2.2), 87
eGFR, mL/min/1.73 m2	63.0 (18.6), 178	55.7 (19.6), 35	61.4 (18.9), 50	66.5 (17.3), 93
Medications, no. of prescriptions	8.8 (4.5), 188	9.5 (3.5), 38	8.8 (5.1), 52	8.5 (4.5), 98
Diuretics, % prescribed	39% (73 of 188)	45% (17 of 38)	38% (20 of 52)	37% (36 of 98)
Drinks, self-reported drinks/day	8.0 (2.7), 168	7.3 (3.0), 31	7.7 (2.8), 48	8.3 (2.3), 90
BMI, kg/m2	25.8 (5.6), 188	26.6 (5.9), 38	26.3 (6.3), 52	25.2 (5.1), 98
BMI <20, %	16% (30 of 188)	11% (4 of 38)	15% (8 of 52)	18% (18 of 98)
Weight, kg	69.0 (17.2), 188	76.0 (20.0), 38	69.2 (17.5), 52	66.1 (15.0), 98
Hemoglobin, g/dL	12.4 (1.5), 170	12.2 (1.0), 34	12.4 (1.8), 47	12.5 (1.4), 89

Note: BMI = body mass index; DRIE = Dehydration Recognition In our Elders; eGFR = estimated glomerular filtration rate; MMSE = Mini-Mental State Examination.

Figure 2.

Percentages of older people living in residential care with current dehydration (20%, serum osmolality > 300 mOsm/kg), impending dehydration (28%, 295–300 mOsm/kg), and who were well hydrated (52%, 275 to <295 mOsm/kg) in the Dehydration Recognition In our Elders (DRIE) study.

Representativeness of Participants

We obtained anonymous data on all residents for 45 (80.4%) of the 56 homes, including 1,425 (78.6%) of 1,812 residents. Of these, 101 were younger than 65 years or had missing age data and so were omitted. DRIE participants were similar in sex ratio but slightly younger than the overall care-home population, with higher BMI and lower likelihood of being undernourished (BMI < 20kg/m²), see Supplementary Table 1.

Characteristics Associated With Serum Osmolality and Dehydration

Characteristics associated (to p < .10) with serum osmolality in univariate linear regression and/or current or impending dehydration in univariate logistic regression (Supplementary Table 2) were very similar to those in the group excluding those with raised/uncertain serum glucose (Supplementary Table 3). These factors associated with osmolality and dehydration included sex, general health factors (eGFR, number of health contacts in past 2 months, number of emergency admissions in past 2 months, diabetic status, from notes [checked with medications list and serum glucose], swollen ankles [current], chronic obstructive pulmonary disease, and arthritis, any type), continence issues (including Barthel Index scores for bowel and bladder continence), cognitive and mental health factors (including MMSE score and MMSE² or MMSE squared, MMSE drawing score, type of consent provided, and staff assessment of dementia level), and use of medications (including those for diabetes, laxatives, loop diuretics, and potassium-sparing diuretics).

Multivariate Analyses

Factors associated with a significantly higher serum osmolality by multivariate linear regression were lower eGFR (signifying worse renal function), lower MMSE (lower cognitive status), diabetic medication use, and not taking potassium-sparing diuretics (Table 2) giving the following regression equation:

Table 2.

Results of Multivariate Regressions for the Full DRIE Data Set (in 188 participants)

Type of Indicator	Linear Regression on Serum Osmolality*		Logistic Regression on Impending Dehydration (90 people have impending or current dehydration)		Logistic Regression on Current Dehydration (38 people have current dehydration)
	Measure	Coefficient (95% CI), p Value	Measure	Odds Ratio (95% CI), p Value	Measure	Odds Ratio (95% CI), p Value
Renal function	eGFR	−0.09 (−0.15 to −0.03), p = .005	eGFR	0.98 (0.96 to 0.99), p = .007	eGFR	0.98 (0.95 to 1.00), p = .045
Cognitive function	MMSE score	−0.37 (−0.56 to −0.18), p < .001	MMSE score	0.93 (0.88 to 0.98), p = .010	MMSE drawing subscore	0.26 (0.11 to 0.65), p = .004
Diabetic status	Uses diabetic medication	6.72 (3.39 to 10.04), p < .001	Diabetic status	3.77 (1.59 to 8.93), p = .003	Use of any diabetic medication	6.77 (2.18 to 21.04), p = .001
Other factors	Potassium-sparing diuretic	−4.93 (−9.09 to −0.77), p = .021	Sex	2.00 (1.01 to 3.97), p = .047	Health care contacts in past 2 months	1.07 (1.01 to 1.13), p = .027

Note: CI = confidence interval; DRIE = Dehydration Recognition In our Elders; eGFR = estimated glomerular filtration rate; MMSE = Mini-Mental State Examination.

*Serum osmolality = 306.0 − (0.086 * eGFR) − (0.37 * MMSE) + 6.72 if uses diabetic medication − 4.93 if uses potassium-sparing diuretics.

$$\begin{array}{l}\text{Serum osmolality = 306.0 }-\text{ (0.086*eGFR) }- \text{(0.37 *MMSE) }\\ \text{+ 6.72 if} \text{uses diabetic medication }-\text{ 4.93 if uses }\\ \text{potassium-sparing diuretic}.\end{array}$$

The regression was also run using MMSE² (as MMSE² was more normally distributed than MMSE), and the equation was similar (Serum osmolality = 302.8 − (0.085*eGFR) − (0.01 *MMSE²) + 6.60 if uses diabetic medication − 4.96 if uses potassium-sparing diuretics).

The pattern of poorer renal function, diabetic status, and poorer cognitive status being associated with higher serum osmolality was largely mirrored in the multivariate logistic regression where MMSE score or subscore, eGFR, diabetic status or use of diabetic medications, male sex, and greater number of recent health care contacts were associated with the odds of dehydration (Table 2).

Each 10-point reduction in eGFR was associated with 20% higher odds of current and impending dehydration and each 10-point MMSE fall was associated with 70% higher odds of impending dehydration. Not being able to draw two intersecting pentagons (part of the MMSE test) was associated with 74% greater odds of current dehydration. Being diabetic was associated with almost quadrupled odds of impending dehydration and use of diabetic medication with sevenfold increase in odds of current dehydration. Men had doubled odds of impending dehydration (compared with women), and every health care contact over the past 2 months was associated with 7% increase in odds of current dehydration.

For analyses omitting participants with raised glucose, we excluded 34 with serum glucose higher than 7.8 mmol/L (>140mg/dl) and 26 without glucose data, leaving 128 people, of whom 22 (17%) had current dehydration and 33 (26%) had impending dehydration (retaining nine participants with diabetes and normal glucose). Univariate analyses are shown in Supplementary Table 3. Multivariate regression (Supplementary Table 4) suggested that factors associated with serum osmolality were renal and cognitive function, use of diabetic medication, and potassium-sparing diuretics, and the regression equation was again similar (Supplementary Table 4 footnote). The reduced analytic power in this smaller data set was clear in the dichotomized logistic regressions, where we struggled to retain statistically significant associations, though eGFR and urinary incontinence were still associated with impending dehydration; cognitive status and renal function were associated with current dehydration.

Fifty participants stated that they felt thirsty before venepuncture. There was no relationship between thirst and serum osmolality (p = .998), see Figure 3, and a receiver operating characteristic plot of thirst and being dehydrated or not gave an area under the curve of 0.47, confirming that thirst is not a good guide to the need to drink in older people.

Figure 3.

Box plot of serum osmolality for those who expressed thirst and those who said they were not thirsty, just before venepuncture.

Discussion

DRIE’s consistent findings (using various statistical models) suggest that cognitive status, renal function, and diabetic status are associated with dehydration in older people. For older people in DRIE, thirst was a poor indicator of need to drink so that drinking must be regulated instead by habit and routine, which are easily disrupted in those with dementia. As renal function declines (45% of DRIE participants had eGFR < 60mL/minute/1.73 m² and 18% eGFR < 45mL/minute/1.73 m²), the ability of older people to conserve fluid declines. Those with diabetes are more likely to experience raised serum glucose, raising serum osmolality, but it is surprising that diabetic medication use is associated with dehydration risk when participants with raised glucose are omitted.

Although DRIE data support the theory of the causation of water-loss dehydration, this was a cross-sectional study. It is quite possible that water-loss dehydration causes poor renal function and poor cognition, and the relationships could work in either or both directions. The relationship between cognitive function and hydration status is potentially highly complex as there is some (but mixed) evidence from children and young adults that dehydration leads to reductions in cognitive performance and mood (36,37). If this is true for older people, then deficits in drinking due to cognitive frailty could lead to dehydration which depresses cognition and thus drinking even further, a vicious circle. We all need to preserve our cognitive function as we age—good hydration may support continued cognitive agility, but more evidence is needed.

A weakness of DRIE was that participants were not totally representative of all care-home residents. DRIE participants were similar in sex ratio but were slightly younger than the overall care-home population and were better nourished. Additionally, for residents who wanted to participate but lacked capacity to give their own consent, we obtained signed consultee agreement for less than 50%, compared with all who gave their own consent (Figure 1), so those with cognitive impairment were relatively under-represented. As cognitive impairment is strongly associated with dehydration, the true proportion of dehydrated care-home residents is probably more than 20%.

Although the group of 188 DRIE participants is still a small sample, this is the first report to our knowledge with a sample of older long-term care residents large enough to examine associations between a good range of health and functional characteristics and serum osmolality or hydration status and to adjust well for confounding.

Serum osmolality, the reference standard for water-loss dehydration in older people, has previously been measured in few, small groups of older people, so assessment of serum osmolality in 188 UK care-home residents is a valuable addition to our understanding and a strength of DRIE as serum osmolality is the best measure of water-loss dehydration. We found only two other care-home studies with serum osmolality measures, both from the United States, and with differing estimates of dehydration prevalence. Gaspar found that 8% of 36 long-term care residents had impending dehydration (0% current dehydration), whereas Stotts found that 19% of 48 nursing home residents at risk for pressure ulcers had current dehydration and a further 44% had impending dehydration (38,39). There are also contradictory findings in hospital settings. In older adults admitted to UK hospitals, levels of dehydration varied from 4% to 58% (40–43) (Supplementary Table 5).

DRIE findings support those of previous small studies that dementia and diabetes are associated with, and potentially risk factors for, dehydration (26,32,33). No previous studies suggest associations with renal function or potassium-sparing diuretics. Our data do not support that age, functional impairments, or nutritional status are associated with hydration status in long-term care residents after adjustment for confounding, but recent health care contacts and incontinence may be important associations, to be assessed in larger studies. The limited size of this data set means that strong but rare predictors of dehydration (such as requiring thickened drinks, a factor for only 8 of DRIE’s participants, see Supplementary Table 2) did not retain high enough p values to be seen in final models. In larger studies, we are likely to find additional strong indicators of dehydration.

Dehydration is associated with increased mortality and disability in older people, but high quality trials to assess effects of interventions to support fluid intake in older people are needed to enable us to formally assess the direction of causation and the health effects of increasing fluid intakes (44,45). Knowing which frail older people are most likely to be, or become, dehydrated will enable future trials to be targeted to promising participants.

Supplementary Material

Please visit the article online at http://gerontologist.oxfordjournals.org/ to view supplementary material.

Funding

This article summarizes independent research funded by the National Institute of Health Research Fellowship programme (grant number NIHR-CDF-2011-04-025, Career Development Fellowship to L.H.). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, the Department of Health, or any other body. The funders played no part in study design after funding was agreed and no part in data collection, analysis, interpretation, writing, or deciding to submit for publication.

Conflict of Interest

The authors declare no conflicts of interest, except that the European Hydration Institute (EHI) supplied plane tickets and paid for 2 night’s hotel accommodation for L.H. to give a talk on dehydration in older people at the EHI’s symposium at the International Congress of Nutrition, Granada, Spain, September 16–18, 2013.