Rosemary Harper

Validating the SF-36 health survey questionnaire: new outcome measure for primary care.

OBJECTIVES--To test the acceptability, validity, and reliability of the short form 36 health survey questionnaire (SF-36) and to compare it with the Nottingham health profile. DESIGN--Postal survey using a questionnaire booklet together with a letter from the general practitioner. Non-respondents received two reminders at two week intervals. The SF-36 questionnaire was retested on a subsample of respondents two weeks after the first mailing. SETTING--Two general practices in Sheffield. PATIENTS--1980 patients aged 16-74 years randomly selected from the two practice lists. MAIN OUTCOME MEASURES--Scores for each health dimension on the SF-36 questionnaire and the Nottingham health profile. Response to questions on recent use of health services and sociodemographic characteristics. RESULTS--The response rate for the SF-36 questionnaire was high (83%) and the rate of completion for each dimension was over 95%. Considerable evidence was found for the reliability of the SF-36 (Cronbachs alpha greater than 0.85, reliability coefficient greater than 0.75 for all dimensions except social functioning) and for construct validity in terms of distinguishing between groups with expected health differences. The SF-36 was able to detect low levels of ill health in patients who had scored 0 (good health) on the Nottingham health profile. CONCLUSIONS--The SF-36 is a promising new instrument for measuring health perception in a general population. It is easy to use, acceptable to patients, and fulfils stringent criteria of reliability and validity. Its use in other contexts and with different disease groups requires further research.

Prolactin and subjective reports of stress in women attending an infertility clinic

Abstract Two samples of patients attending an Infertility Clinic for the first time were asked to complete an Adjective Check List and the State-Trait Anxiety Inventory respectively, on arrival at the Clinic. Prolactin was estimated from blood samples taken during the same visit, and before examination by a gynaecologist. Significant positive correlations were obtained between measures of ‘stress’, ‘state’ anxiety and prolactin. The incidence of omissions and errors was also related to ‘state’ anxiety. This anticipatory response of prolactin, correlated with subjective reports, confirms the lability of prolactin to psychological factors.

Validating the SF-36: Authors' reply

0-100, which seems unlikely for a score derived from responses to questions concerning subjective symptoms. They say they have used the method of Bland and Altman,2 but the results given-the mean difference and the distribution of the differences-are not good measures of reliability. Moreover, Bland and Altman proposed that the limits of agreement should be used to assess repeatability. For example, twice the standard deviation of the differences gives a range within which 95% of the differences will lie. A mean difference ofzero implies that there is no consistent trend affecting test and retest results (all the patients getting better, for example). It does not imply that the measure is reliable. The table gives some hypothetical data to show how you can get a mean difference of 0 and still have widely different test and retest scores, because differences in different directions cancel each other out whatever their absolute size.

THE AHEAD STUDY: MANAGING ANTICOAGULATED PATIENTS WHO SUFFER HEAD INJURY

Objectives & Background Existing practice in emergency departments (ED) in the UK for managing anticoagulated patients after blunt head trauma is variable and based on limited evidence. We aimed to determine the head injury complication rate within this group of patients and identify risk factors associated with a poor outcome. Methods A prospective observational multi-centre study enrolled patients taking warfarin who attended 33 emergency departments in England and Scotland after blunt head trauma. ED attendance data and patient reported outcomes were collected over an 18-month period from October 2011. Head injury complication was defined by head injury-related death, neurosurgery resulting from injury, clinically-significant CT head scan or re-attendance with significant head injury complications. Factors including neurological status (Glasgow Coma Score, GCS), level of anticoagulation (INR) and neurological symptoms were entered into multivariate logistic regression analyses as predictors of a poor outcome. Results A total of 3566 patients were enrolled; anonymised clinical data was submitted for 99.1% of patients. The age range was 18 to 101 years (median 81 yrs, IQR 12), 48.7% were men. Mean initial INR was 2.67 (SD 1.34, IQR 1.1), 80.5% patients had a GCS of 15 and 68.1% did not report any associated neurological symptoms. 59.3% of patients had a CT head scan with a significant head injury-related finding identified in 5.6%. Reversal therapy was given to 5.3% of patients, only 19 (0.5%) patients underwent neurosurgery and 42 (1.2%) patients died of a head injury-related death. The overall rate of head injury complication was found to be 6.1%. Multivariate logistic regression modelling found GCS <13 (OR 12.7; 95% CI 2.41 to 67.2), vomiting (OR 3.09; 95% CI 1.71 to 5.61) and loss of consciousness (OR 2.53; 95% CI 1.54 to 4.15) to be significantly associated with a poor outcome. In univariate analyses an INR >4 was associated with an increased risk of an adverse outcome, however it made no significant contribution to the multivariable models. Conclusion This is the largest cohort of anticoagulated head injury patients ever reported. The head injury complication rate was 6.1% which correlates well with previous findings. INR was not found to be associated with a poor outcome however GCS <13, vomiting and loss of consciousness were identified as significant risk factors for an adverse outcome in anticoagulated patients with a head injury.