The Care Programme Approach (CPA; Department of Health, 1990) has been subject to criticism based on lack of evidence for its usefulness (Reference Tyrer, Morgan and Van HORNTyrer et al, 1995) and the bureaucratic burden associated with its implementation (Reference KingdonKingdon, 1998, Reference MarshallMarshall, 1999). There has also been wide variation in the numbers of patients placed on CPA registers, which is not explained by variation in need (Reference Bindman, Beck and GloverBindman et al, 1999). This may be partly explained by differing CPA policies and procedures. In old age psychiatry services there can be variation between trusts bound by the same CPA criteria within the same health authority (Reference Philpot, Sheehan and ReevesPhilpot et al, 1998; Reference Wallace and BallWallace & Ball, 1998). Variation in the use of the register within the same trust has also been observed (Reference Philpot, Banerjee, Johnson, Ramsey and ThornicroftPhilpot & Banerjee, 1997), despite the criteria for inclusion on the register being clearly set down in trust policy. To investigate this variation further we examined the adherence to CPA criteria and the factors affecting the likelihood of a patient's inclusion on the CPA register in three distinct geographical areas of the trust.

The study

The old age psychiatry service of the Bethlem & Maudsley Trust served the London Borough of Croydon (4 consultants), the eastern part of Lambeth (1 consultant) and the southern part of Southwark (3 consultants). In the first survey the case notes of all patients placed on the CPA register between January and March 1998 were reviewed. Demographic and clinical data, including diagnosis, were recorded as well as the presence of the criteria listed below. Register patients were compared with a similar number of ‘control’ patients whose cases were active at the same time but who were not placed on the register. Cases were matched for sex, age and borough of home address.

The results of the first audit survey were discussed at local audit meetings and disseminated throughout the service. The second survey included all patients registered between October 1998 and March 1999. On this occasion, five control patients were randomly selected from each of nine geographical sectors. The data were collected in the same way as in the first survey.

Twenty-one patients were registered during the first survey period (3 months) and 56 during the second (6 months). Results were compared with 22 and 45 control patients, respectively.

Findings

At the first survey there were no differences between CPA patients and controls in terms of age, sex, proportion from ethnic minorities, those living alone or unmarried and diagnosis. However, CPA patients were more likely to have involvement of social services (P=0.001), a history of previous admissions (P=0.0001), a longer period of contact with the service (P=0.004) and greater psychotropic drug prescription (P=0.026) than controls. At the second survey only social services involvement (P=0.021) and the proportion of patients living alone (P=0.031) differentiated the two groups.

Table 1 shows the proportion of CPA patients and controls fulfilling each class of criteria. With the exception of ‘category (a)’ criteria during the first study, CPA patients were more likely than controls to fulfil each of the category requirements for registration. If strictly applied, values for ‘per cent fulfilling (a) and (b) and/or (c)’ should have been 100% and 0% for CPA patients and controls, respectively. Treating the CPA policy criteria as the gold standard, Table 2 shows how sensitivity, specificity and misclassification rate differed between the three boroughs within the trust at the two survey times. The only significant changes occurred in sensitivity (i.e. the proportion of cases fulfilling the CPA criteria that were actually registered). Sensitivity fell in Croydon but rose in South Southwark. In East Lambeth there were trends suggesting an increase in sensitivity and a fall in misclassification rate. At the second survey misclassification rates for individual consultants varied between 9 and 63%. Table 2 also shows the average monthly patient registration rates during each audit. The patient's borough was an important factor in registration rate, practice varying between surveys particularly in Croydon and South Southwark.

Table 1. Proportion of registered patients and controls meeting Care Programme Approach (CPA) criteria

	First survey			Second survey
	CPA	Control	P 1	CPA	Control	P 1
Number of cases	21	22		56	45
Per cent with major mental illness (a)	100	82	0.059	73	44	0.003
Per cent with at least one risk factor (b)	81	36	0.003	86	51	0.0002
Per cent fulfilling (a) and (b)	81	36	0.003	64	36	0.004
Per cent where section 117 applies (c)	33	0	0.004	20	0	0.007
Per cent fulfilling (a) and (b) and/or (c)	81	0	0.003	73	33	0.006

Table 2. Adherence to Care Programme Approach criteria and registration rate in the three geographical areas served by the trust

Survey	Croydon		East Lambeth		South Southwark
	First	Second	First	Second	First	Second
Sensitivity (%)	78	291	57	78	50	912
Specificity (%)	67	91	50	88	100	50
Misclassification rate (%)	29	44	46	18	33	27
Patients registered per month per 100 000 elderly in catchment area during survey periods	7	0.2	2	1	1	4

Discussion

Our results confirm the variability in the use of the CPA register and the application of the defining criteria in an old age psychiatry service. During the discussion following presentation of the first survey results it became apparent that some consultants believed that no clinical benefits to patients were derived from the use of the CPA register and that in one borough social workers would only assess those patients placed on the register. Reaction to the first audit survey varied. The rate of CPA registration by one consultant team rose dramatically to include nearly all patients referred to the service, while registration in one other team virtually ceased.

In our study we did not seek to determine whether the clinical care of individual patients was in any way affected by the CPA policy or whether the variation in practice had a bearing on the quality of care given. Slavish adherence to guidelines does not necessarily guarantee quality of service (Reference Marshall, Gray and LockwoodMarshall et al, 1997; Reference Schneider, Carpenter and BrandonSchneider et al, 1999). However, failure to fully apply clinical policies such as the CPA has lead to criticism of psychiatric services in a number of recent serious incident inquiries (Reference Kelly and DevauxBaroness Scotland of Asthal et al, 1998).

Health service policies are written to reduce variations and to eliminate unacceptable omissions in clinical practice. Where policies are developed in negotiation with clinicians, as was the case with the policy examined here, it is reasonable to expect closer adherence than was found in this study. It is possible that weaknesses in one aspect of a clinician's practice reflects problems elsewhere. Audits of routine matters such as the CPA may be one method of ensuring acceptable practice within the framework of clinical governance.