Data Collection

This study was a part of a larger study to assess the impact of incentive payments on general practice performance one year before (2004) and one year after (2005) the introduction of the general medical service contract in 2004 [14]. The patient inclusion criteria were any patient aged 18 or over, with diagnosed hypertension and at least one record of systolic BP > 150 or diastolic BP > 90, prior to 1 April 2003 and 2005 respectively. Data were extracted from the patients' electronic and paper records. Sampling from 2003 and 2005 was conducted independently. Between 20 and 40 randomly selected eligible patients at each practice were contacted for permission to examine their full records. All participants gave informed consent. Patient's age, gender, smoking status, year of diagnosis and postcodes were retrieved from the notes.

Patient and Practice deprivation calculation

The Office for National Statistics (ONS) determined the Index of Multiple Deprivation (IMD) in 2004 [15] by combining a number of indicators chosen to cover a range of economic, social and housing issues into a single deprivation score for each small area. These small areas are called LSOA (Lower Super Output Area), of which there are 32,482 in England [16]. Each LSOA consists of about 1,500 people within a defined geographical locality. This allows each area to be ranked relative to one another according to their level of deprivation [15]. Nationally IMD scores for each LSOA vary from 0.37 in the least deprived area to 85.46 in the most deprived one.

Each patient's index of multiple deprivation (IMD) 2004 score was determined by mapping his/her postcode to the relevant LSOA [15, 16]. Patients were categorised into three groups according to their IMD scores in comparison to the rest of the sample. To give almost equal numbers in each group, patients with IMD scores of less than 11.6 were grouped in the least deprived group, those between 11.6 and 18.9 in the middle group and those with an IMD of 19 or more in the most deprived group.

The practice deprivation score was the weighted mean of the IMDs for the postcodes of all registered patients. Practice deprivation scores ranged from 8%-85% of the full range of national practice deprivation scores. Practices were classified according to the national deprivation rank for general practices into three groups. The practices with a total deprivation rank below 30% of the national deprivation rank were classed as most deprived and those above 65% were classed as least deprived.

Data Analysis

The primary outcome measurement for each participant was the achievement of each quality indicator with a binary (yes/no) score. For each indicator, the percentage of eligible patients for whom the indicator was achieved was calculated.

Associations between indicator achievement and patient and practice characteristics were evaluated using logistic regression models. The outcome variable was quality indicator achievement, and the explanatory variables were patient and practice characteristics. Patient characteristics were: age group (< 63 years old, 63-73, or > 73), gender, smoking status, deprivation group, time point (2003 or 2005), year of diagnosis (1967-1997 or 1998-2005). Practice characteristics were practice size and deprivation group. If 5,000 patients or less were registered in the practice, it was classed as small. Those practices with more than 10,000 patients were classed as large practices.

Firstly, we performed univariate analyses for each quality indicator against patient and practice characteristics separately. We also carried out a subgroup deprivation analysis by time; dividing the sample into two groups (2003, 2005). Secondly, we did a multivariate analysis for each indicator including all possible explanatory variables, we eliminated in a stepwise fashion those variables that were not statistically significant, until only statistically significant (p < 0.05) exposure variables remained in the model. Data analyses were conducting using STATA (Version 8 SE, Texas US).

Association between deprivation and achievement-univariate analysis

The results were similar whether deprivation was derived from patient postcode area deprivation or from practice deprivation. All results presented below refer to patient postcode area deprivation rather than practice deprivation.

For non-incentivised quality indicators, the achievement of indicators did not vary significantly by deprivation for 8 out of 10 indicators. For the remaining two indicators, achievement was significantly higher in the more deprived areas. Achievement ranged from 7% to 85% in the least deprived third, compared to 24%-93% in the most deprived third.

Subgroup univariate analysis was then performed grouping patients according to time point into 2003 or 2005 group. For incentivised indicators in 2003, there was no difference in achievement of 3 out of 4 indicators by area deprivation, with the remaining one indicator showing significantly higher achievement in more deprived groups. In 2005 there were no differences by deprivation for incentivised indicators. For non-incentivised indicators, achievement of 2 of 10 indicators was significantly higher in more deprived groups in both 2003 and 2005.

Association between deprivation and achievement after adjustment for other factors-multivariate analysis

The primary aim of this analysis was to estimate the effect of area deprivation. In addition, we identified a number of significant associations between other characteristics and indicator achievement, which will be summarised here. Men were twice as likely as women to have an assessment of peripheral vascular disease (nonQOF 2) or diabetes (nonQOF 3) (Tables 3 & 5). Participants aged over 73 years old were almost four times more likely to be assessed for peripheral vascular disease compared to those under the age of 63 (p < 0.001) and they are also more likely to be assessed for presence of protein in urine (Tables 3 & 5). As reported previously, nearly all indicators improved over time, with greater achievements in incentivised indicators (13). Large practices were less likely to assess peripheral vascular disease (odds ratio 0.47, 95% confidence interval 0.28-0.77) or check protein in urine (0.69, 0.49-0.98) (Table 5). No association was found between the quality of care and patient smoking status.

Summary of main findings

The achievement of quality indicators for hypertension either did not vary with geographic deprivation, or was higher in patients from more deprived localities. There were few differences in inequalities before and after the implementation of incentives, although there was a general trend for quality to improve over time, and some indicators reached a ceiling of high achievement in all deprivation categories.

Additionally, we found that using patients' postcode-based deprivation scores gave similar results to using practices' area deprivation scores, derived from the weighted mean of relevant IMDs to the postcodes of all practice's registered patients.

Strengths and limitations of the study

This study has a number of strengths. To the best of our knowledge, this is the first study to assess the relationship between deprivation and non-incentivised indicators of care for hypertension. All the indicators are evidence-based and validated by independent panels including general practitioners (Table 1). Data were collected by hand searching both the electronic and paper patient records using clear criteria. Credit was given to any mention of the care even if it was not fully documented.

There was an adequate response rate of 76%, and the responders were similar with respect to age and sex to non-responders. Therefore, we think it unlikely that response bias has had a substantial affect on the results, although it is possible that there were differences in the level of deprivation between responders and non-responders. The practices in this study were broadly representative of the English national range of the socioeconomic deprivation [14]. Most of the quality indicators in this study referred to processes of the health care rather than outcomes. Process measures have the following advantages over outcome measures for assessing the quality of health care. There are many causes of changes in health status other than health care, and there are many problems in adequately adjusting outcomes for differences in case mix [17]. Processes are also more sensitive measures of quality than outcomes, and more clearly linked to any action that should be taken to improve quality [18]. Moreover, process measures are currently the basis of the Quality and Outcomes Framework and other pay for performance schemes, and therefore are clearly relevant to those schemes and this paper [19, 20].

This study assessed patient area deprivation using patient postcodes, rather than practice postcodes which have been used in previous studies [7–9]. Although using patients' postcode-based deprivation scores is a more accurate measure of deprivation than simply using the practice postcode deprivation, the patients' socioeconomic status might vary despite sharing the same postcode. Ideally further studies should evaluate the correlation of quality of care with patient's income and level of education, but this information is difficult to obtain.

Since 2004 the GP contract has been reviewed, and in 2006 the two hypertension indicators related to smoking were amalgamated with other smoking indicators onto a single smoking domain [19, 20].

Limitations include that we assessed recorded care, and it is possible that the care was delivered without being recorded. However, recording is an essential component of quality in a team-based approach to chronic disease management [21]. The study was also based on a small number of practices and patients, due to the practical constraints of collecting data manually from patient records. We categorised participants into thirds in terms of area deprivation. We set the IMD range of these subgroups to give almost equal numbers in each group rather than comparative to the national IMD range. This is because the analysis was set to evaluate the health inequalities within the sample. Only three bands of deprivation were compared. As there is a chance that the differences in quality lie in the bottom quintile, not the bottom tertile, we repeated the analysis using different break points for deprivation, with no substantial changes to the results, and so are confident that our results are not driven by the methods chosen. Nevertheless, a study which focuses on more extreme deprivation may find different results, and this would be an interesting area for further research.

We did not assess the effect of severity of disease, and all patients with hypertension who met our inclusion criteria were eligible. The ethnicity of the patients was not considered in the analysis, as 95-97% of the Norfolk population is classified as 'White British' [22].

Whilst we set out to investigate the relationship between deprivation and quality, our secondary analyses included approximately 70 other univariate analyses comparing baseline characteristics of patients with our varied quality indicators. We used a conventional p-value of < 0.05 to determine statistical significance, but could have used a p-value of p < 0.01 or less in view of the multiple statistical testing. This would have produced a less nuanced but more consistent result that deprivation had no significant effect on achievement of any of the indicators.

Comparison with existing literature

Our findings are consistent with those reported by Strong et al, who assessed quality of care in 38 general practices. They found either no association between indicator achievement and deprivation, or better care in more deprived areas [9]. Equally, 8,515 practices were assessed by Ashworth et al [8], who reported slightly lower achievement of incentivised indicators for hypertension in 2004/5 in the most deprived areas compared to the least deprived areas, with the near disappearance of this gap by 2006/7.