|Home | About | Journals | Submit | Contact Us | Français|
In England, patients can choose to have their NHS elective care delivered by private (or ‘independent sector’) providers or by the NHS. Providers are paid a fixed tariff for each type of procedure. Our objectives were to compare NHS providers with private treatment centres in terms of (a) the quality of data coding and (b) patient complexity.
We compared elective patients aged 18 years and over treated in the NHS and private sectors using the Hospital Episode Statistics (HES) data for 2007–2008. The absence of diagnostic information was used as a measure of data quality. We analysed differences in complexity for each of the 30 Healthcare Resource Groups (HRGs) that together account for 78% of coded private treatment centre activity. Statistical significance was assessed at the 1% level.
Hospitals and treatment centres.
Patient complexity was assessed by four characteristics: age; number of diagnoses; number of procedures; and income deprivation of residential area.
NHS providers treated almost 7 million adult elective patients in 2007–2008. Fewer than 100,000 patients were treated by private providers (1.3% of elective activity). Less than 1% of NHS patients lacked diagnostic information compared to 36% of patients treated by private providers. For the top 30 HRGs, NHS patients had significantly (p<0.01) higher levels of co-morbidity, underwent more procedures and were more likely to come from deprived areas compared with patients treated by private providers. Although patients treated in private settings tended to be younger, the difference was not statistically significant.
Some private companies provide poor quality data. In general, the NHS is treating more complex patients than private providers. If complexity drives costs, then a fair reimbursement system would require higher payments for NHS providers.
Involving the private sector in the NHS was viewed as crucial to the government's objectives of offering patients a greater choice of provider and reducing waiting times. Most private (or ‘independent sector’) providers are treatment centres that specialize in a small number of high volume procedures, such as hip replacements or cataract removals, and that avoid taking on complex operations.
The Independent Sector Treatment Centre (ISTC) programme commenced in 2003. Under wave 1, care was commissioned from 26 ISTCs and an ophthalmic chain of 12 cataract centres.1 In 2007, a further 15 ISTCs were commissioned under phase 2. The total cost of these two commissioning phases was estimated at £2.7bn.2
The first wave of private treatment centres were paid 11% above the price for equivalent treatment in the NHS, even though they had lower staffing costs, offered inferior pensions and claimed that their ‘private sector’ mentality made them more efficient.3 Moreover, they were paid irrespective of whether contracted volumes were delivered.3 These generous contractual terms, intended to encourage market entry and cover start-up costs, prompted questions about whether the Department of Health was securing value for money in commissioning from the private sector.3 The second phase of private treatment centres received less generous terms.
It was intended that, once fully established, the ISTC programme would provide some 250,000 elective procedures and 1.5m diagnostic assessments per year.4 But both elective and diagnostic activity has fallen far short of these targets.5 Our previous research identified only 30,000 elective patients treated by the private sector in 2005–2006 and 66,000 in 2006–2007, barely 0.5% and 1%, respectively, of the 6.7 million elective patients treated by the NHS in each of these periods.6
Moreover, it has been difficult to determine precisely what types of elective procedures have been provided because the quality of data from some private treatment centres has been so poor. Although contracts mandated the provision of data for the Hospital Episode Statistics,4 incentives or sanctions associated with these data returns were absent. The first purpose of this paper is to assess, on a company-by-company basis, the amount of activity and quality of coding by the private sector in 2007–2008.
As well as treating fewer patients than expected, the private sector may be treating patients of lower complexity than those treated by the NHS. Under new contractual arrangements, termed the Extended Choice Network, ISTCs will be subject to Payment by Results whereby all providers are paid the same ‘national tariff’ for particular types of operation – defined as Healthcare Resource Groups (HRGs). This arrangement assumes that NHS and private providers treat similar patients. Our analysis of data for 2005–2006 and 2006–2007 suggested that private treatment centres were admitting easier cases within some HRGs.6,7
This finding may reflect an insufficient refinement of version 3.5 HRG categories, which could not adequately differentiate patients with different healthcare requirements. Since our earlier research, the HRG classification system has been substantially overhauled: version 4 contains around 1400 HRGs compared to the 610 categories in version 3.5. It was hoped that the greater degree of granularity offered by the new HRG system would eliminate case-mix differences between NHS and private providers among patients classified to the same HRG. The second purpose of this paper is to assess whether this hope has been realized.
We analysed the data in the Hospital Episode Statistics (HES) for 2007–2008. The HES contain details about every patient funded by the NHS in England whether treated by NHS or private providers. There have been a number of takeovers and mergers of ISTCs in recent years, and our analysis takes ownership status as applied at the start of the 2007–2008 financial year. We excluded patients treated in Care Trusts or Primary Care Trusts. Patients in HES are defined as ‘finished consultant episodes’ and each patient was assigned an HRG code using the new (fourth) version of HRGs by the NHS Information Centre.
We undertook two analyses. First, we compared the quality of HES coding by NHS providers (hospitals and treatment centres) with coding by private treatment centres. We follow the Healthcare Commission4 in considering the absence of information about the patient's primary diagnosis as indicative of poor coding. If this information is absent, the patient is allocated to the ‘uncoded’ HRG, UZ01Z. We consider the number of patients allocated to this HRG as a proportion of each provider's total activity.
Second, we compared the complexity of patients treated in the NHS and private sector in terms of four characteristics: their age; the number of diagnoses recorded per patient; the number of procedures the patient underwent; and the income deprivation of the patient's residential area. The analysis was restricted to elective patients aged 18 years and over, who were classified to any one of the 30 HRGs that accounted for the highest volume of private treatment centre activity. These 30 HRGs cover 78% of total coded private sector activity.
For each HRG and each characteristic, we conducted t-tests to determine the statistical significance of observed differences between patients treated in the NHS and private sector. As multiple tests are performed on the same patients, statistical significance was assessed at the 1% level (i.e. p<0.01) to reduce the risk of making a type I error (i.e. observing a difference when there is none).
To present our results, we created forest plots using RevMan5.8 Forest plots are commonly used in meta-analysis to compare results across different clinical trials,9 but in this case, our comparisons are across the 30 HRGs. For each HRG, the difference between patients treated in the NHS and private sector is represented as a rectangle (showing the mean difference) and a horizontal line through each rectangle (showing the 99% confidence interval). If the confidence interval intersects the vertical axis (at 0), there is no statistically significant difference between patients treated in the two settings for the HRG and characteristic in question.
We also estimated an overall weighted mean difference (WMD) and associated confidence interval (CI). The WMD is calculated by summing the mean differences across all HRGs, weighted by the number of patients in each HRG using a random effects model. This approach implies more conservative claims about statistical significance than the alternative fixed effects model.9,10 The WMD is shown as a black diamond at the bottom of each figure. If the diamond intersects the vertical axis, the overall difference is not statistically significant.
NHS providers treated 6,966,436 adult elective patients in 2007–2008. Only 93,457 were treated by private providers, some 1.3% of all elective activity ( Table 1) and far fewer than the annual 250,000 elective procedures originally anticipated from the ISTC programme.
The second row of Table 1 shows that less than 1% of NHS patients were allocated to the ‘uncoded’ HRG, UZ01Z, compared to 36% of patients treated by private providers. However, there are marked differences in coding across private providers ( Figure 1). For instance, UK Specialist Hospitals achieves an equivalent level of coding to NHS providers, with only 1% of records being uncoded. In marked contrast, Capio,a which runs 10 ISTCs, performed poorly on coding quality: 97% of its 15,827 patient records contained insufficient diagnostic information to allow assignment to an HRG.
Table 1 lists the top 30 HRG codes by volume of activity undertaken by private providers. The most frequent categories of HRGs were H codes (Musculoskeletal System), F codes (Digestive System), B codes (Eyes and Periorbita) and L codes (Urinary Tract and Male Reproductive System). There are marked differences across HRGs in the proportionate split in activity between the private sector and the NHS. Although accounting for just 1.3% of overall elective activity in 2007–2008, the private sector undertook substantially higher proportions of certain types of operation. For example, private providers accounted for 23.9% of activity coded to BZ03Z (Non-Phacoemulsification Cataract Surgery); 24.0% of activity coded to FZ16Z (Very Major Procedures for Gastrointestinal Bleed); and 10% of activity coded to HB21C (Major Knee Procedures for non Trauma).
The results of the HRG analyses are shown in Figures 2–5. For each indicator of complexity, the figures show the difference between the mean values for private providers relative to the NHS, together with a pooled value across all 30 HRGs – the weighted mean difference (WMD) – shown at the bottom of the figure. For example, the top row in Figure 2 shows that the mean difference in the number of diagnoses for the HRG ‘BZ02Z’ (Phacoemulsification Cataract Extraction and Lens Implant) was −1.15, indicating that patients treated by private providers had on average 1.15 fewer diagnoses than patients treated by the NHS. The black diamond, showing the WMD, does not cross the vertical axis. This means that across all 30 HRGs, patients treated in the private sector had significantly fewer recorded diagnoses than patients treated in the NHS.
Overall, the analyses provide some clear messages. First, significantly fewer diagnoses were recorded for patients treated by private providers compared with NHS providers (WMD: −0.59 [99% CI −0.79 to −0.39]) and these differences were systematic across all HRGs (Figure 2). Second, patients treated by private providers generally underwent significantly fewer procedures compared with NHS patients (WMD: −0.66 [99% CI −0.81 to −0.51]) (Figure 3). Third, patients treated in the private sector were on average one year younger than patients treated in the NHS. However, there was no systematic pattern across all HRGs and the difference was not statistically significant (WMD: −0.99 [99% CI −2.66 to −0.67]) (Figure 4). Finally, for all but one HRG,b private providers were less likely than the NHS to serve patients from deprived residential areas (WMD: −1.56% [99%CI −2.21% to −0.92%]) (Figure 5).
We used subgroup analysis to examine the data by HRG category. For H codes (Musculoskeletal System) and for B codes (Eyes and Periorbita), findings were similar to those of the full analysis. However, F code (Digestive System) patients treated by the NHS were significantly more complex than their private sector counterparts onall four indicators, including age. In contrast, the subgroup analysis of L codes (Urinary Tract and Male Reproductive System) showed that patients treated in the private sector had more diagnoses, but were otherwise not significantly more complex than those treated in the NHS.
There were two principal findings from the study. First, data coding quality is generally high in NHS providers, but private providers vary in their coding accuracy: some providers code their data as well as NHS organizations, but others perform very poorly. Second, our analysis of national data strongly suggests that NHS organizations are treating a more complex case-mix than their private sector counterparts. In three out of four indicators of patient complexity – the number of diagnoses, number of procedures and income deprivation – NHS organizations were found to be treating significantly more complex patients than private providers. On average, NHS organizations were also treating slightly older patients although the difference was not statistically significant.
We adopted a methodologically simple comparison of patient complexity based on weighted mean differences. An alternative approach would have been to estimate regression models to assess the influence of each patient characteristic on the probability of being treated in one setting or another. This would have allowed us to estimate the effect of each characteristic ‘conditional’ upon the influence of the other characteristics, rather than the ‘unconditional’ effects presented here. The conditional and unconditional effects will differ if the variables are highly correlated but, for our data, the correlations are very low (all less than r = 0.12). In the interests of statistical parsimony, we chose the simpler analytical approach.
In summarizing our results, we pooled data across HRGs, despite the presence of high levels of ‘heterogeneity’ (or variation). We report heterogeneity in each figure, shown by the I-squared statistic (I2=100% is the highest possible level). While it is generally desirable to avoid combining ‘apples’ and ‘oranges’ (different HRGs), this approach is acceptable if the aim is ‘to contribute to a wider question about fruit’.9 In summarizing across HRGs, our purpose is to identify whether we can answer broad questions about the types of patients treated in different settings. The advantage of using forest plots is that they clearly show both the pooled value and the individual mean differences for each of the 30 HRGs.
Our findings on case-mix suggest that early concerns about the ISTC programme have yet to be resolved. Based on data for 129 cataract patients seen in an NHS hospital, Barsam and colleagues reported an increase in case-mix complexity after the opening of a nearby ISTC.11 A study by Browne compared 769 patients treated in six ISTCs with 1895 patients treated in 20 NHS providers and found that those treated in ISTCs had less complex conditions.12 The authors caution against extrapolation because few ISTCs participated and the sample of patients was small. Nonetheless, these studies questioned whether the private sector treatment centre programme was providing value for money. Previous research, using HRG 3.5 data from 2006–2007, found that private providers were treating less complicated patients.7 Our present study, which considers all elective patients treated in 2007–2008, shows that the problem persists,despite improvements to the HRG categorization system.
Additional concerns have been raised about the ISTC programme, including the relationship with other NHS providers, implications for clinical training, levels of activity delivered, health outcomes and patient experience.5,13–15 Our study did not assess the health outcomes – these data are not yet routinely reported at a national level – but one small study found tentative evidence that ISTCs perform better for cataract surgery and hip replacement, whereas the NHS achieves better outcomes for hernia repair.12
If the data submitted by healthcare providers are of poor quality, this limits their usefulness for monitoring of the type and standard of care provided to NHS patients and prevents proper scrutiny of the use to which NHS funds are put. Unlike NHS providers, private providers are currently paid for uncoded activity, thanks to the generous contracts used to initiate market entry. Although contracts mandated good coding practice, this has apparently not been enforced.2 When these contracts end, private providers will be paid the same price as the NHS and this price will depend on the HRG assigned to the patient. This will provide a stronger financial incentive to improve coding quality, but other measures may be necessary to secure coding compliance by companies with inadequate returns.
An unintended consequence of the start-up contracts is that they may undermine patient choice. There is anecdotal evidence that Primary Care Trusts (PCTs), facing large deficits because of under-utilized capacity for which they must pay, are putting pressure on GPs to refer patients to private providers.16 However, this would appear to conflict with the aim of giving patients freedom to choose their provider for elective care. This phenomenon underscores the need for joined-up policy-making that aligns incentives within regulatory and payment systems, and that carefully monitors the systems for potential unintended consequences.
Payment by Results is predicated on the assumption that, for any given HRG, patients treated by NHS and private providers have the same care requirements. Our previous analysis, based on an earlier version of HRGs, found that patients treated in the NHS were significantly more complex than those seen by private sector treatment centres. It was hoped that version 4 HRGs would minimize the differences by classifying patients with greater precision. But our analysis has shown that the problem persists, with the NHS continuing to care for more complex patients than private providers. If data quality improves, future research can determine the full extent of this problem and establish whether these observed differences in patient characteristics drive costs. If so, then a fair reimbursement system would require higher payments for NHS providers.
Competing interests None declared
Funding The project was funded by the Department of Health in England as part of a programme of policy research
Ethical approval Not applicable
Contributorship AS conceived the idea for the paper and AM developed it; RV analysed the data; AM and AS wrote the first draft and produced the figures; AS and RV refined the draft
Reviewer Azeem Majeed
The authors acknowledge the assistance provided by Nathan Abbotts, Adriana Castelli, Susan Devlin, Henry Forster, Mauro Laudicella,Eileen Robertson and Peter Sivey. The views expressed are those of the authors and are not necessarily those of the Department of Health
Chart shows coded and uncoded activity for each ISTC group, ordered by percentage of coded activity (from highest [left] to lowest [right])
Activity: measured as Finished Consultant Episodes (FCEs)
[N]: number of treatment centres within each ISTC group
aCapio treatment centres were acquired by Ramsay Healthcare in November 2007
bLB33Z: Vasectomy Procedures