Time trend analysis and variations in prescribing lipid lowering drugs in general practice

The first trial to show that patients with coronary heart disease treated with lipid lowering drugs gained a survival advantage was published in November 1994.¹ Other similar trials that used hydroxymethyl glutaryl coenzyme A (HMG-CoA) reductase inhibitors, or statins, have subsequently confirmed these results (Long-term Intervention with Pravastatin in Ischaemic Disease (LIPID) Study Group, 70th scientific sessions of the American Heart Association, Florida, November 1997).² Despite this, consistent failure occurs in the implementation of these findings.³ This study of primary care in South East Thames investigated variation between practices in the use of lipid lowering drugs and examined how prescribing has changed over time between different health authorities since 1990.

Subjects, methods, and results

Four health authorities were chosen to represent the range of demographic variables (including age, ethnic group, and social status) found in this region. These authorities were Bexley and Greenwich; Lambeth, Southwark, and Lewisham; East Kent; and East Sussex.

Time trend analysis was performed with prescribing analysis and cost (PACT) data from the Prescription Pricing Authority. Aggregate section trends for lipid lowering drugs were obtained for each health authority from April 1990 to September 1996. Because of boundary changes data from East Sussex were available only from April 1992. The raw data were adjusted for differences in population size and age stratification by calculating health authority net ingredient cost of lipid lowering drugs per patient prescribing unit per month. We used the convention adopted by the pricing authority to give a weighting of three prescribing units for every patient aged 65 years and over and unity for all younger patients. Values for population size and number of patients aged 65 years and over were included in the raw PACT data.

Cross sectional analysis of prescribing was performed with practice profile sections for lipid lowering drugs. As general practitioners may use different methods of repeat prescribing—for example, monthly, two monthly, or quarterly—we collected data from a 3 month period, July to September 1996. These were used to calculate practice net ingredient cost per patient prescribing unit per quarter (with the same weighting system as above). Again list size and number of patients aged 65 years and over were included in the PACT data. Results were analysed with SPSS statistical software.⁴

Changes in prescribing of lipid lowering drugs over time in the four health authorities were described by a single model with an initial linear phase followed by an exponential phase (superimposed on the time trend shown in the figure). The change point from linear to exponential was calculated for each health authority by using the least squares technique and by minimising the residual sum of squares with respect to the change point. This was found to be closely related to the publication date of the Scandinavian simvastatin survival study.¹

One way analysis of variance showed that differences between health authority spending on lipid lowering drugs were highly significant during both phases of the model (variance ratio (F) P<0.00001). During the exponential phase the time taken for authority spending on lipid lowering drugs to double varied from 16 months (Bexley and Greenwich) to 28 months (East Sussex).

Results of the cross sectional analysis show that prescribing of lipid lowering drugs by individual general practices is highly variable. Practices in Lambeth, Southwark, and Lewisham prescribed significantly fewer lipid lowering drugs than practices in other health authorities (F=56; P<0.0001). Similarly, practices in Bexley and Greenwich prescribed significantly fewer lipid lowering drugs than those in East Kent and East Sussex. Even within a single health authority prescribing varied up to 60-fold between practices, and a 98-fold variation existed across the South East Thames region as a whole.

Comment

Since November 1994 prescribing of lipid lowering drugs increased exponentially in all health authorities studied, but the rate of change varied widely. Use of these drugs also varied greatly between individual general practices. We suggest that the recent increase is linked to the availability of research evidence, but further studies are needed to determine if variation in prescribing between authorities and practices reflects differences in clinical need.

Figure.

Prescribing of lipid lowering drugs in general practice in four health authorities, 1990-6. Linear phase r², exponential phase r², and change points were 0.94, 0.98, and November 1994 for Lambeth, Southwark, and Lewisham; 0.92, 0.97, and December 1994 for Bexley and Greenwich; 0.88, 0.97, and February 1995 for East Kent; and 0.84, 0.93, and January 1995 for East Sussex. 4S is Scandinavian simvastatin survival study¹