Design: Cross sectional study of the rate of retirement due to ill health by age, sex, and length of service. Principal diagnoses by age and length of service were also compared.

Setting: Four public and two private large employers in the United Kingdom.

Main outcome measures: Rates of retirement on the grounds of ill health by age, sex, and length of service of employees contributing to pension schemes.

Results: Rates of ill health retirement varied from 20 to 250 per 10,000 contributing members, and in two organisations the rate varied geographically within the same organisation. In the two organisations that provided data by sex, women retired at a greater rate than men under age 40 and over age 50. In four organisations the modal age or length of service coincided with enhancements in benefits. In the four that provided information on diagnoses, musculoskeletal and minor psychiatric illnesses were the most common reasons for retirement.

Conclusion: The granting of ill health retirement benefits may not be determined by illness. There is a need for some employers and pension schemes to improve their processes for granting benefits. Doctors should be wary of conflicts of interest and work to guidelines when they advise pension schemes about the merits of an application for benefits.

Dudley Priority Health NHS Trust,
Central Clinic,
Dudley,
West Midlands DY2 7BX
C J M Poole, consultant occupational physician

Design: Prospective cross sectional study.

Setting: Acute medical admissions ward of a teaching hospital.

Subjects: 71 randomly selected patients admitted with acute dyspnoea.

Main outcome measures: Sensitivity and specificity of each investigation and logistic regression analysis of each variable in identifying left ventricular systolic dysfunction.

Results: Clinical assessment in this cohort of patients with severe dyspnoea was generally sensitive (sensitivity 81%). Patients were divided into three groups on the basis of clinical assessment. In the first group (37 patients) the diagnosis of systolic dysfunction was clear, in the second (22) it was in doubt, and in the third (12) it was unlikely. The sensitivity of clinical assessment in identifying left ventricular systolic dysfunction was 81% and the specificity was 47%. The specificity of diagnosis was improved by electrocardiography (69%) and chest radiography (92%). Logistic regression analysis showed that isolated pulmonary crepitations were a comparatively poor predictor of left ventricular systolic dysfunction (chi-2 =10.215, P=0.0014) but that a full clinical examination had reasonable predictive value (chi-2 =24.82, P<0.00001). The combination of clinical assessment and chest radiography improved the accuracy of diagnosis (chi-2 =28.08, P<0.00001), as did the combination of clinical assessment and electrocardiography (chi-2 =32.41, P<0.00001).

Conclusion: Clinical assessment in patients admitted with acute dyspnoea is comparatively accurate. Patients with abnormal results on chest radiography, electrocardiography, and clinical examination have a high likelihood of having left ventricular systolic dysfunction. Echocardiography contributes little more to the diagnosis in these patients and may be more efficiently directed towards patients in whom the diagnosis is still in doubt after clinical assessment, chest radiography, and electrocardiography.

Correspondence to: Dr N D Gillespie,
Department of Medicine (Section of Ageing and Health),
Ninewells Hospital and Medical School,
Dundee DD1 9SY.

Departments of Cardiology,
Clinical Pharmacology, and Epidemiology,
Ninewells Hospital,
Dundee DD1 9SY
Neil D Gillespie, research fellow, department of cardiology
Graeme McNeill, consultant cardiologist
Terence Pringle, consultant cardiologist
Simon Ogston, statistician, department of epidemiology
Allan D Struthers, professor of clinical pharmacology
Stuart D Pringle, consultant cardiologist

Design: Review of medical records in general practices to identify hypertensive patients followed up by assessment of the pre-educational and post-educational management of interventions.

Setting: Six general practices in north west London picked at random within defined criteria of geography and size.

Subjects: 740 hypertensive patients managed by 25 different general practitioners.

Main outcome measures: Improved level of care in terms of better diagnosis by having at least three blood pressure readings before the start of drug treatment, better level of recordings of lifestyle parameters as shown by the level of recordings of body mass index and total lipid values, and better control of blood pressure and harm minimisation as shown by the level of recordings of urea and electrolyte values.

Results: Improvement was noted in the level of recordings of body mass index, total lipid concentrations, and urea and electrolyte values but not in better diagnosis or blood pressure control.

Conclusion: Clinical behaviour of general practitioners can be changed by peer review but more complex behavioural changes which require the cooperation of the patients and cognitive actions by the general practitioners need further investigation.

Department of Clinical Pharmacology and Therapeutics,
Charing Cross and Westminster Medical School,
Chelsea and Westminster Hospital,
London SW10 9NH
Mahendra Mashru, North Thames clinical research fellow in general practice
Ariel Lant, professor

Correspondence to: Dr Mahendra Mashru,
19 King Edward's Road,
Ruislip,
Middlesex HA4 7AG.