Old people are commonly receiving diuretics on admission to hospital. Diuretics are recognized as a risk factor for electrolyte disturbances; controversy exists about the relative risks of different combinations (in particular, co-amilozide [Moduretic]). We recorded the drug history and serum electrolytes in 1000 consecutive admissions to a geriatric hospital, and examined the relative prescribing rates of various diuretics in the community. Full results were obtained in 929 patients. A history of diuretic prescription was present in 353 (38%) of the patients; the mean serum sodium in this group (95% CI 136.0-137.1 mmol/l) was lower than in the 586 not prescribed diuretics (137.1-137.9 mmol/l). The difference was small but statistically significant (95% CI difference = 0.3-1.6 mmol/l; P less than 0.01). Hyponatraemia (serum sodium less than 130 mmol/l) was not significantly commoner in the 41 patients prescribed co-amilozide than in patients prescribed other diuretics. In general patients prescribed potassium-retaining diuretics had a lower serum sodium than the others. There was a significant positive correlation between the serum potassium and the log [serum urea] (r = 0.26, P less than 0.001) and a weak negative correlation existed between sodium and potassium (r = -0.14; P less than 0.001). There was an association between the prescription of potassium-retaining diuretics and a higher serum potassium; also an association between the prescription of a loop or thiazide diuretic and a lower serum potassium. These interactions were shown by multiple regression analysis to be independent and additive.(ABSTRACT TRUNCATED AT 250 WORDS)