Risk assessment in primary prevention of coronary heart disease: randomised comparison of three scoring methods

That lipid lowering with statins benefits even those at low risk of coronary heart disease is no longer open to question. The challenge now is for clinicians to strike a balance between what is desirable, affordable, and achievable. As serum total cholesterol concentration alone poorly predicts cardiovascular risk, alternative methods of risk assessment have been proposed. We compared the ability of general practitioners and practice nurses to interpret three of these methods. We chose the revised Sheffield table,¹ the New Zealand guidelines,² and the joint British chart³ because all three included age, sex, smoking and diabetes status, blood pressure, and ratio of total cholesterol to high density lipoprotein cholesterol as part of their risk assessment.

Subjects, methods, and results

All 37 general practices in Dumfries and Galloway, in Scotland, were randomised to receive the three risk scores in different sequences, each with the same set of 12 case histories. A self nominated general practitioner and nurse in each practice were each asked whether coronary risk exceeded 3% per year (Sheffield table), whether it exceeded 30% over 10 years (joint British chart), or whether cardiovascular risk exceeded 20% over five years (New Zealand guidelines) for each case history. These thresholds were chosen to reflect current practice.⁴,⁵ Doctors and nurses also rated each guideline for ease of use and preference, using scales from 1 to 5 (5=easiest or most preferred).

Accuracy, ease of use, and preference were compared for doctors and nurses separately, first with Freidman's test overall and then with Wilcoxon's signed rank tests on the differences for each subject for pairs of guidelines. P values reported are unadjusted for multiple comparisons, but the results stand after correction with the Bonferroni method.

Two practices did not have a practice nurse. In another practice the same nurse did not score all three guidelines, and so the results were excluded from the analyses of ease of use and preference. In all, 33/37 doctors and 22/35 nurses scored at least 10 of 12 case histories correctly when using the Sheffield table; corresponding numbers for the New Zealand guidelines were 37 and 33 respectively and for the joint British chart 36 and 34 respectively. There were no significant differences between the three scores for doctors, whereas accuracy among nurses was significantly poorer (P<0.001) with the Sheffield table than with each of the other two guidelines (table). Only 6 doctors and 6/34 nurses gave the Sheffield table a high preference rating (4 or 5). More doctors and nurses gave high preference scores for the New Zealand guidelines (26 doctors and 25 nurses) and for the joint British chart (23 and 25) (P<0.001 for the Sheffield table compared with each of the other two guidelines for both doctors and nurses). Similar results were found for ease of use (table).

Comment

Of these three risk assessment methods, nurses are more likely to interpret correctly the New Zealand guidelines and joint British chart, and both general practitioners and nurses not only find these two methods easier to use but also prefer them to the Sheffield table.

The main strength of our study was that a named general practitioner and nurse within every practice in Dumfries and Galloway completed a formal assessment of each of the three risk scores. A possible limitation is that the study was confined to a single health board. We have no reason to believe, however, that general practitioners and nurses in Dumfries and Galloway are unrepresentative of their colleagues elsewhere in Scotland and the United Kingdom or that the responses would have been different had we assessed the risk scores during clinical contacts.

We have shown that cardiovascular risk assessment by tables and charts based on the Framingham equation is acceptable to both general practitioners and nurses. The results of our study favour the New Zealand guidelines and the joint British chart, the latter of which may be the more suitable for use in primary care. The continuous scale for systolic pressure facilitates assessment of blood pressure and the risk chart is also available as a computer program.

Table.

Comparison of three methods of risk assessment for coronary heart disease among 37 general practitioners and 35^* practice nurses. Values in parentheses are approximate 95% confidence intervals

Test	Overall P value	New Zealand guidelines v Sheffield table			Joint British chart v Sheffield table			Joint British chart v New Zealand guidelines
		Median difference	P value		Median difference	P value		Median difference	P value
Doctors:
Accuracy	0.36	0 (−0.5 to 0.5)	0.48		0 (0 to 0.5)	0.34		0 (0 to 0.5)	1.00
Preference	<0.001	1.5 (1 to 2)	<0.001		1.5 (1 to 2)	<0.001		0 (−0.5 to 0.5)	0.66
Ease of use	<0.001	1.5 (1 to 2)	<0.001		1.5 (0.5 to 2)	<0.001		0 (−0.5 to 0.5)	0.63
Nurses:
Accuracy	0.002	1 (0.5 to 2)	<0.001		1.5 (0.5 to 2.5)	<0.001		0 (−0.5 to 0.5)	0.73
Preference	<0.001	2.5 (1.5 to 3)	<0.001		2 (1.5 to 2.5)	<0.001		0 (−0.5 to 1)	0.55
Ease of use	<0.001	2 (1 to 2.5)	<0.001		2 (1 to 2.5)	<0.001		0 (−0.5 to 1)	0.65

The overall P value is a Friedman test; the pairwise comparisons are Wilcoxon's signed rank tests on the median differences between matched pairs. 

^*34 for analyses of preference and ease of use (see text). 

Editorial by Jackson