Water-loss dehydration results from drinking too little fluid. It is common in older people and associated with increased risk of many health problems. We wanted to find out whether simple tests (like skin turgor, dry mouth, urine colour and bioelectrical impedance) can usefully tell us whether an older person (aged at least 65 years) is drinking enough. Within the review we assessed 67 different tests, but no tests were consistently useful in telling us whether older people are drinking enough, or are dehydrated. Some tests did appear useful in some studies, and these promising tests should be re-checked to see whether they are useful in specific older populations. There was sufficient evidence to suggest that some tests should not be used to indicate dehydration. Tests that should not be used include dry mouth, feeling thirsty, heart rate, urine colour, and urine volume.
There is limited evidence of the diagnostic utility of any individual clinical symptom, sign or test or combination of tests to indicate water-loss dehydration in older people. Individual tests should not be used in this population to indicate dehydration; they miss a high proportion of people with dehydration, and wrongly label those who are adequately hydrated.
Promising tests identified by this review need to be further assessed, as do new methods in development. Combining several tests may improve diagnostic accuracy.
There is evidence that water-loss dehydration is common in older people and associated with many causes of morbidity and mortality. However, it is unclear what clinical symptoms, signs and tests may be used to identify early dehydration in older people, so that support can be mobilised to improve hydration before health and well-being are compromised.
To determine the diagnostic accuracy of state (one time), minimally invasive clinical symptoms, signs and tests to be used as screening tests for detecting water-loss dehydration in older people by systematically reviewing studies that have measured a reference standard and at least one index test in people aged 65 years and over. Water-loss dehydration was defined primarily as including everyone with either impending or current water-loss dehydration (including all those with serum osmolality ≥ 295 mOsm/kg as being dehydrated).
Structured search strategies were developed for MEDLINE (OvidSP), EMBASE (OvidSP), CINAHL, LILACS, DARE and HTA databases (The Cochrane Library), and the International Clinical Trials Registry Platform (ICTRP). Reference lists of included studies and identified relevant reviews were checked. Authors of included studies were contacted for details of further studies.
Titles and abstracts were scanned and all potentially relevant studies obtained in full text. Inclusion of full text studies was assessed independently in duplicate, and disagreements resolved by a third author. We wrote to authors of all studies that appeared to have collected data on at least one reference standard and at least one index test, and in at least 10 people aged ≥ 65 years, even where no comparative analysis has been published, requesting original dataset so we could create 2 x 2 tables.
Diagnostic accuracy of each test was assessed against the best available reference standard for water-loss dehydration (serum or plasma osmolality cut-off ≥ 295 mOsm/kg, serum osmolarity or weight change) within each study. For each index test study data were presented in forest plots of sensitivity and specificity. The primary target condition was water-loss dehydration (including either impending or current water-loss dehydration). Secondary target conditions were intended as current (> 300 mOsm/kg) and impending (295 to 300 mOsm/kg) water-loss dehydration, but restricted to current dehydration in the final review.
We conducted bivariate random-effects meta-analyses (Stata/IC, StataCorp) for index tests where there were at least four studies and study datasets could be pooled to construct sensitivity and specificity summary estimates. We assigned the same approach for index tests with continuous outcome data for each of three pre-specified cut-off points investigated.
Pre-set minimum sensitivity of a useful test was 60%, minimum specificity 75%. As pre-specifying three cut-offs for each continuous test may have led to missing a cut-off with useful sensitivity and specificity, we conducted post-hoc exploratory analyses to create receiver operating characteristic (ROC) curves where there appeared some possibility of a useful cut-off missed by the original three. These analyses enabled assessment of which tests may be worth assessing in further research. A further exploratory analysis assessed the value of combining the best two index tests where each had some individual predictive ability.
There were few published studies of the diagnostic accuracy of state (one time), minimally invasive clinical symptoms, signs or tests to be used as screening tests for detecting water-loss dehydration in older people. Therefore, to complete this review we sought, analysed and included raw datasets that included a reference standard and an index test in people aged ≥ 65 years.
We included three studies with published diagnostic accuracy data and a further 21 studies provided datasets that we analysed. We assessed 67 tests (at three cut-offs for each continuous outcome) for diagnostic accuracy of water-loss dehydration (primary target condition) and of current dehydration (secondary target condition).
Only three tests showed any ability to diagnose water-loss dehydration (including both impending and current water-loss dehydration) as stand-alone tests: expressing fatigue (sensitivity 0.71 (95% CI 0.29 to 0.96), specificity 0.75 (95% CI 0.63 to 0.85), in one study with 71 participants, but two additional studies had lower sensitivity); missing drinks between meals (sensitivity 1.00 (95% CI 0.59 to 1.00), specificity 0.77 (95% CI 0.64 to 0.86), in one study with 71 participants) and BIA resistance at 50 kHz (sensitivities 1.00 (95% CI 0.48 to 1.00) and 0.71 (95% CI 0.44 to 0.90) and specificities of 1.00 (95% CI 0.69 to 1.00) and 0.80 (95% CI 0.28 to 0.99) in 15 and 22 people respectively for two studies, but with sensitivities of 0.54 (95% CI 0.25 to 0.81) and 0.69 (95% CI 0.56 to 0.79) and specificities of 0.50 (95% CI 0.16 to 0.84) and 0.19 (95% CI 0.17 to 0.21) in 21 and 1947 people respectively in two other studies). In post-hoc ROC plots drinks intake, urine osmolality and axillial moisture also showed limited diagnostic accuracy. No test was consistently useful in more than one study.
Combining two tests so that an individual both missed some drinks between meals and expressed fatigue was sensitive at 0.71 (95% CI 0.29 to 0.96) and specific at 0.92 (95% CI 0.83 to 0.97).
There was sufficient evidence to suggest that several stand-alone tests often used to assess dehydration in older people (including fluid intake, urine specific gravity, urine colour, urine volume, heart rate, dry mouth, feeling thirsty and BIA assessment of intracellular water or extracellular water) are not useful, and should not be relied on individually as ways of assessing presence or absence of dehydration in older people.
No tests were found consistently useful in diagnosing current water-loss dehydration.