Skip Navigation


Journal of Public Health Advance Access originally published online on February 29, 2008
Journal of Public Health 2008 30(2):186-193; doi:10.1093/pubmed/fdn016
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
30/2/186    most recent
fdn016v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Austin, D.
Right arrow Articles by Pell, J. P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Austin, D.
Right arrow Articles by Pell, J. P.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2008, Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved

Hospital and operator variations in drug-eluting stent use: a multi-level analysis of 5967 consecutive patients in Scotland



David Austin, Clinical Research Fellow in Cardiology1

Keith G. Oldroyd, Consultant Cardiologist and Honorary Senior Lecturer2

Alex McConnachie, Statistician3

Rachel Slack, Scottish Cardiac Registers Co-ordinator1

Hany Eteiba, Consultant Cardiologist and Honorary Senior Lecturer4

Andrew D. Flapan, Consultant Cardiologist5

Kevin P. Jennings, Consultant Cardiologist6

Robin J. Northcote, Consultant Cardiologist7

Alistair C. H. Pell, Consultant Cardiologist8

Ian R. Starkey, Consultant Cardiologist9

Jill P. Pell, Professor of Public Health1
1 Section of Public Health and Health Policy, University of Glasgow, 1 Lilybank Gardens, G12 8RZ Glasgow, UK
2 Western Infirmary, Glasgow, UK
3 Robertson Centre for Biostatistics, University of Glasgow, Glasgow, UK
4 Glasgow Royal Infirmary, Glasgow, UK
5 Edinburgh Royal Infirmary, Edinburgh, UK
6 Aberdeen Royal Infirmary, Aberdeen, UK
7 Victoria Infirmary, Glasgow, UK
8 Monklands Hospital, Airdrie, Lanarkshire, UK
9 Western General Hospital, Edinburgh, UK

Address correspondence to Jill P. Pell, E-mail: j.pell{at}clinmed.gla.ac.uk

Objective To determine whether drug-eluting stent (DES) use varies among Scottish hospitals, and the extent to which any variations are explained by differences between operators, patients and lesions.

Methods Multi-level analysis of consecutive patients treated with percutaneous coronary intervention (PCI) between April 2005 and March 2006 in Scotland, using the Scottish Coronary Revascularization Registry.

Results A total of 38 operators performed 5967 PCI procedures on 8489 lesions. Crude level of DES use was 47.6%, and the results varied among hospitals (range 30.6–61.8%, {chi}2 = 341.6, P < 0.0001). There was significant between-operator variation in the null model. This was attenuated by the addition of hospital as a fixed effect. Nonetheless, the final model demonstrated significant between-operator variability [{sigma}2 = 0.486 (0.249–0.971)] and between-hospital variation, after case-mix adjustment.

Conclusions Within Scotland, marked variation existed among hospitals in the use of DES. Operator was the most important factor at patient level, and hospital of treatment, rather than case-mix, was the most important modifier of between-operator variation. Patient selection for DES is complex and may contribute to much of the variations demonstrated. Consensus criteria would provide more detail than is included in current guidance, may aid decision-making for individual patients, reduce opportunity costs and ensure equity of access.

Keywords: coronary artery disease, drug-eluting stents, multi-level model, percutaneous coronary intervention, practice variation


   Introduction
TOP
 Introduction
Methods
 Results
 Discussion
 Conclusion
 Conflict of interest
 Acknowledgements
 References
 
Drug-eluting stents (DES) have been rapidly assimilated into routine cardiology practice for the treatment of coronary heart disease. Compared with conventional bare-metal stents (BMS), DES reduce restenosis1,2—a complication of percutaneous coronary intervention (PCI) that increases the risk of recurrent symptoms and repeat procedures. DES do not reduce death or myocardial infarction and are more expensive than BMS; therefore, controversy exists regarding their cost-effectiveness.3,4 In the UK, guidelines were developed at an early stage in the evaluation of DES, with cost-effectiveness a significant consideration in the final report.5 DES use was recommended for treating vessels <3 mm or lesions >15 mm. Previous studies have demonstrated differences in DES utilization among hospitals, largely attributed to differences in healthcare funding within countries.6,7 However, no study has examined clinical practice within the UK healthcare system, or determined whether practice variation extends to operator level. This is particularly important in light of the significant opportunity costs associated with rapid expansion of an expensive new technology.

In comparing practice among hospitals, it is imperative to adjust for case-mix differences. Analysis of stent choice during PCI is further complicated by the fact that patients may have more than one lesion treated during a procedure. Some factors, such as age and co-morbidity, are patient characteristics while others, such as angiographic features, act at lesion level. Furthermore, unmeasured factors may influence individual operator practice, independent of case-mix and hospital factors. Conventional multi-variate analysis assumes independence of analysis units (in this case lesions), and may thus overestimate variations at higher levels of the hierarchy (in this case patients, operators and hospitals). Multi-level modelling, also known as random effects modelling, incorporates the data hierarchy into the statistical model, allowing a more accurate comparison.8–10 In addition to producing odds ratios for fixed effects, multi-level analysis quantifies the residual, or ‘unexplained’, variation at the different levels of the hierarchy. The aim of our study was to determine whether DES use varied among hospitals and operators, and whether variation persisted after adjustment for case-mix.


   Methods
TOP
 Introduction
 Methods
Results
 Discussion
 Conclusion
 Conflict of interest
 Acknowledgements
 References
 
Data source and study population
Since 1997, the Scottish Coronary Revascularization Register has routinely collected detailed information prospectively on all PCIs performed in Scottish hospitals. Data entry is completed by a combination of clinical and administrative staff according to a pre-defined set of standardized data definitions, and centrally collated to form the Scottish register. The database is annually validated for completeness and consistency. Information collected includes demography, co-morbidity, clinical presentation and procedural details. In Scotland, very few PCIs are performed in the private sector. Therefore, our analysis was restricted to the seven publicly funded National Health Service (NHS) hospitals that undertake PCI. We included consecutive patients treated with PCI within these hospitals over a 1 year period from April 2005. Informed consent to collate and use data is obtained from patients prior to coronary revascularization. All patient and operator data are stripped of unique identifiers and, for the purposes of reporting, hospitals are not identified by name.

Definitions
Hypertension was a binary variable and defined as blood pressure >140/90 mmHg or treatment with antihypertensive therapy. Hyperlipidaemia was defined as total cholesterol concentration >5.2 mmol/l or treatment with a lipid-lowering agent. Diabetes was defined as either type I or type II diabetes mellitus. Urgent cases were defined as revascularization undertaken during the index admission and emergency cases as revascularization undertaken within 24 h of admission/referral.

Statistical analysis
DES use varied over the 12-month period analysed; therefore, quarters of the year were included as a categorical variable to account for changes over time. Hospital was included as a fixed effect, rather than the top level of the hierarchy, for two reasons: with only seven hospitals, there were too few to use hospital as the top level of the hierarchy in a multi-level model;11 also, one-third of operators practised at more than one hospital, and a multi-level model fully incorporating this cross-classified structure would have been overly complex.

We defined a binary outcome based on DES use in each lesion. If both DES and BMS were employed in the same lesion, DES use was recorded. We compared crude proportions of DES use by hospital using a {chi}2 test. All patient and lesion co-variates listed in Table 1 were tested univariately for their association with DES use. All factors significantly associated with DES use at the 5% level were included in subsequent multi-level analyses. All variables were treated as categorical, with the exception of age which was continuous. No significant co-linearity was noted.


View this table:
[in this window]
[in a new window]

  		Table 1 Patient and lesion characteristics by hospital

 
A multi-level logistic regression model was constructed according to the data hierarchy, with random variation permitted at three levels: operator, patient and lesion. This three-level structure is referred to as the null model, to which hospital, patient and lesion fixed effects were subsequently added. First, hospital was entered as a fixed effect (model 1). Then, patient and lesion fixed co-variate effects (model 2) were assessed. We chose to add patient- and lesion-fixed effects grouped as ‘case-mix’, because more than two-thirds of patients (67.3%) are treated for single lesions. Finally, hospital, patient and lesion fixed effects were added within the same model (model 3). All models were adjusted for change over time. We report the variance estimate (posterior median [2.5–97.5 percentile range)] for between-operator and between-patient variation for each model. Variance estimates are not produced for the lowest level in the hierarchy (lesion). In the interests of brevity, odds ratios for patient and lesion fixed effects are only shown for Model 3 (i.e. the final fully fitted model). Hospital estimates are shown as unadjusted, and for Model 3 with P-value calculated with a {chi}2 test. All multi-level analyses were fitted using Markov Chain Monte Carlo (MCMC) estimation procedure, with a 5000 iteration ‘burn-in’ phase and 50000 iteration chain length. Iteration histories were checked visually to assess mixing. Bayesian Deviance Information Criterion (BDIC) and deviance (MCMC) statistics are shown as an indicator of the change in model fit; reductions in these values indicate an improvement.12 Post hoc analyses were subsequently performed to investigate further the operator-level variation.

The descriptive analyses were performed using SPSS v14.0 software (SPSS Inc, Chicago, IL, USA). The multi-level analyses were performed using MLwiN v2.02 (Centre for multi-level Modelling, University of Bristol, UK).


   Results
TOP
 Introduction
 Methods
 Results
Discussion
 Conclusion
 Conflict of interest
 Acknowledgements
 References
 
Between April 2005 and March 2006, PCI was performed on 8863 lesions. We excluded 218 lesions because no device was recorded, 128 because the procedure was abandoned and 28 because the operator code was missing. The remaining 8489 lesions equated to 5967 PCI procedures undertaken by 38 operators in the seven hospitals.

Overall, DESs were used in 47.6% of the lesions treated. However, this varied significantly by hospital (Fig. 1, {chi}2 = 341.6, P < 0.0001) with the crude percentage use ranging from 30.6 to 61.8%. The paclitaxel-eluting Taxus stent (Boston Scientific Corp., Natick, MA, USA) accounted for 80.7% of the DES deployed, and the sirolimus-eluting Cypher stent (Cordis Corp, Miami, FL, USA) for 18.6%.


Figure 1
View larger version (10K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 1 Unadjusted hospital estimates and three-level adjusted hospital estimates for drug-eluting stent use.

 
There were no statistically significant differences between hospitals with regard to sex and diabetic status, but a large number of patient and lesion characteristics were found to vary by hospital of treatment (Table 1). The null three-level random effects model demonstrated significant variation between operators and between patients (Table 2). Adding the hospital-fixed effect reduced the operator-level variation by >40% ({sigma}2 = 0.792 to {sigma}2 = 0.445), though patient-level variation increased slightly (Model 1, Table 2). When all univariate case-mix predictors of DES use were added, there were only minor alterations in variance estimates (Model 2, Table 2), though model fit was improved (indicated by reduced BDIC and deviance statistics for Model 2, Table 2). The final model demonstrated variance estimates similar to Model 1; significant ‘unexplained’ variation at both operator and patient level remained, though further improvements in model fit were noted (Model 3, Table 2). Between-hospital variation is represented in Fig. 1. Variation persisted even after adjusting for the data hierarchy and case-mix ({chi}2 = 22.1, P = 0.001), though, as expected, the confidence intervals widened.


View this table:
[in this window]
[in a new window]

  		Table 2 Three-level random effects models for DES use

 
Table 3 demonstrates the fixed effects estimates for case-mix variables included in Model 3 (the final adjusted multi-level analysis). DES use at patient level was associated with younger, male patients and with more stable presentations of coronary artery disease. The most powerful predictors of DES use were lesion factors, however. Left main and left anterior descending PCI, the presence of restenosis, longer lesions, smaller vessels and less tortuous vessels were all associated with an increased probability of DES use. Increased DES use over the 12 months studied is accounted for by the time period variable.


View this table:
[in this window]
[in a new window]

  		Table 3 Patient and lesion fixed effects within the final three-level model

 
To analyse the effect of operator on patient-level variation, a post hoc two-level model (patient and lesion) was run without the effect of operator. Between-patient variation increased to {sigma}2 = 3.959 (3.202–4.813), an increase of 25%. A further analysis of below and above median operator volume as a fixed effect revealed no association with DES use, and no difference to operator-level variance estimates within the full multi-level hierarchy.

As a supplement to the findings of the multi-level models, DES use for the highest and lowest operator quartiles were compared (Table 4). Within high and low operator use groups, differential DES use by clinical category is evident. Between high and low operator groups, DES use was statistically no different among patients treated for restenosis and left main coronary artery lesions. Marked differences between low and high operator DES use were noted among all other sub-groups, though for lesions at high baseline risk of restenosis (e.g. ≤2.75 mm and lesions >20 mm length), between group differences were relatively smaller.


View this table:
[in this window]
[in a new window]

  		Table 4 Low and high operator DES use, by clinical indication

 

   Discussion
TOP
 Introduction
 Methods
 Results
 Discussion
Conclusion
 Conflict of interest
 Acknowledgements
 References
 
Main findings from this study
Two key features emerge from this study. Within a single, centrally funded healthcare system, significant hospital variation in DES use existed even after appropriate multi-level adjustment. Secondly, between-operator decision-making varied significantly, independent of case-mix, and was the most important determinant of patient-level variation. Operator-level variation was greatly reduced, however, when hospital was introduced as a factor, suggesting local factors heavily influence clinical practice

Previous studies from the USA and Italy concluded that differences in the adoption of DES within these healthcare systems were likely to be due to financial considerations, e.g. private versus academic and government institutions, and the proportion of patients insured.6,7 Such distinctions are less relevant within the UK NHS, where all treatments are publicly funded. Furthermore, Scotland has a single national Coronary Heart Disease taskforce and only three regional planning groups responsible for the PCI service. Yet, this study demonstrated marked hospital differences in the use of DES even after adjustment for operator practice. However, the variations in hospital practice observed may still be attributable to financial considerations through differential local priority setting (‘postcode prescribing’), stent availability or integrated care pathways, as well as less tangible influences on group clinical practice.

Significant between-operator variation could exist for a number of reasons. The impact of National Institute for Health and Clinical Excellence guidance on some operators is reflected in the higher use of DES in patients, for example, with long coronary stenoses.5 Clinical practice is more complex, however, and includes many more lesion types and patient groups. The important clinical variables and their relative influence on practice are listed in Table 3. Many patient sub-groups commonly encountered have been studied in only small numbers, or excluded from randomized, controlled trials (RCTs). Notable examples include patients treated for non-ST elevation myocardial infarction and multi-vessel PCI. Interventional cardiologists may therefore differ in their interpretation of the effectiveness of DES in these patients. Some may choose to extrapolate the results of RCTs on selected patients to groups at high baseline risk of restenosis and therefore infer a higher absolute benefit from DES. Others may restrict DES to patients who would have fulfilled the entry criteria for RCTs, or strictly follow the existing guidelines. These difficulties in applying published studies and guidelines to clinical practice are likely to have contributed to the operator-variation we demonstrated.

Our study period largely predates a recent clinical controversy surrounding late stent thrombosis with DES.13–16 This is a serious but rare complication that does not appear to compromise patient outcomes in those studied in RCTs.17–19 However, late stent thrombosis seems to be more common among patients who would not have been included in pivotal RCTs.20,21 For resource-limited healthcare systems, this introduces a potential paradox. Treating complex patients at high baseline risk of restenosis may be more cost-effective; however in such patients the RCT evidence of safety is generally weaker. This conflict has further complicated clinical decision-making in individual patients.

What is already known on this topic?
DES reduce restenosis and, therefore, the need for repeat procedures following PCI.1,2 Consequently, they have been rapidly and widely adopted into clinical practice. Clinical guidelines have been developed, and concluded that DES were clinically effective, but in many common situations, the absolute benefit derived did not justify their use.5

What this study adds
Utilization of DES varies widely both between hospitals and, independently, between operators, within a single centrally funded healthcare system. Hospital variation may exist within the UK NHS for a number of reasons, and had a marked impact on operator practice. The existence of such striking variation also raises questions regarding the applicability and relevance of current guidelines to individual patients. Such a finding is important given the costs associated with DES use and the volume of PCI procedures in the UK.

Strengths and weaknesses of this study
The strengths of our study include comprehensive case ascertainment, collection of detailed procedural co-variates key to clinical decision-making, and the representation of actual clinical practice. Though the current study was limited to Scotland, the findings are likely to reflect current practice within the UK NHS. In addition, the principles of operator and hospital variation during the adoption of an expensive new technology have wider relevance to many other countries that have similar financial restrictions to the UK.

Because our study was observational, the case-mix of patients undergoing PCI varied significantly by hospital. Observed case-mix differences are likely to reflect differing baseline patient populations and the selection of other forms of treatment such as coronary artery bypass grafting, the threshold for which may vary by hospital. An individual operator may also be selected by those referring on the basis of clinical factors (e.g. lesion type) as well as non-clinical characteristics (e.g. threshold for intervention); both are likely to affect the case-mix of individual operators. We attempted to adjust for resultant differences in case-mix within the study design; however, as with all observational studies, unmeasured or unknown clinical confounders may play a role. We were able to show that operator volume did not account for the residual operator-level variation demonstrated. However, we had little further data on operator characteristics that may ‘explain’ why individual clinicians' choices vary beyond case-mix differences. This aspect merits further study.

Finally, we demonstrated a large amount of ‘unexplained’ patient-level variation that was largely unaffected by the addition of case-mix co-variates. This variation is entirely attributable to the patients who received multi-lesion PCI (around one-third of the total population). Our findings suggest that lesion-level rather patient-level factors determine DES use, with little correlation between the two. Therefore, it is likely that these factors contributed to the large amount of random variation at patient level.


   Conclusion
TOP
 Introduction
 Methods
 Results
 Discussion
 Conclusion
Conflict of interest
 Acknowledgements
 References
 
Our study demonstrated wide variations in DES use between hospitals and operators. It is unclear whether the inequalities in use of DES reflect under-utilization by some operators and hospitals or over-utilization by others; both have important consequences. It could be argued that the existing clinical guidelines do not reflect the complexities faced in clinical practice, nor are they sufficiently detailed to assist the calculation of risk and benefit for individual patients. Comprehensive consensus criteria would provide a clinically relevant starting-point to define appropriate DES use for individual patients, in order to aid clinical judgement, maximize benefit, minimize opportunity costs and ensure equity of access.


   Conflict of interest
TOP
 Introduction
 Methods
 Results
 Discussion
 Conclusion
 Conflict of interest
Acknowledgements
 References
 
Dr D. Austin and Prof. J. P. Pell have received research funding from Boston Scientific. Dr KG Oldroyd has received speaker and consultancy fees from Boston Scientific, Medtronic and Cordis J & J.


   Acknowledgements
TOP
 Introduction
 Methods
 Results
 Discussion
 Conclusion
 Conflict of interest
 Acknowledgements
References
 
We are grateful to the data co-ordinators and administrators at each contributing site.


   References
TOP
 Introduction
 Methods
 Results
 Discussion
 Conclusion
 Conflict of interest
 Acknowledgements
 References
 

  1. Stone GW, Ellis SG, Cox DA, et al. One-year clinical results with the slow-release, polymer-based, paclitaxel-eluting TAXUS stent: the TAXUS-IV trial. Circulation (2004) 109:1942–7.[Abstract/Free Full Text]
  2. Moses JW, Leon MB, Popma JJ, et al. Sirolimus-eluting stents versus standard stents in patients with stenosis in a native coronary artery. N Engl J Med (2003) 349:1315–23.[Abstract/Free Full Text]
  3. Bagust A, Grayson AD, Palmer ND, Perry RA, Walley T. Cost effectiveness of drug eluting coronary artery stenting in a UK setting: cost-utility study. Heart (2006) 92:68–74.[Abstract/Free Full Text]
  4. Thomas M. Are drug eluting stents really worth the money? Heart (2006) 92:5–7.[Free Full Text]
  5. National Institute for Health and Clinical Excellence (NICE). Guidance on the use of coronary stents http://www.nice.org.uk/page.aspx?o=TA071guidance (3 December 2007, date last accessed).
  6. Grilli R, Guastaroba P, Taroni F. Effect of hospital ownership status and payment structure on the adoption and use of drug-eluting stents for percutaneous coronary interventions. CMAJ (2007) 176:185–90.[Abstract/Free Full Text]
  7. Rao SV, Shaw RE, Brindis RG, et al. Patterns and outcomes of drug-eluting coronary stent use in clinical practice. Am Heart J (2006) 152:321–6.[CrossRef][Web of Science][Medline]
  8. Austin PC, Jack V, Alte DA. Comparing hierarchical modeling with traditional logistic regression analysis among patients hospitalized with acute myocardial infarction: Should we be analyzing cardiovascular outcomes data differently? Am Heart J (2003) 145:27–35.[CrossRef][Web of Science][Medline]
  9. Gupta M, Chang WC, Van de Werf F, et al. International differences in in-hospital revascularization and outcomes following acute myocardial infarction: A multilevel analysis of patients in ASSENT-2. Eur Heart J (2003) 24:1640–50.[Abstract/Free Full Text]
  10. Chang WC, Midodzi WK, Westerhout CM, et al. Are international differences in the outcomes of acute coronary syndromes apparent or real? A multilevel analysis. J Epidemiol Commun Health (2005) 59:427–33.[Abstract/Free Full Text]
  11. Snijders TAB Sampling. Multilevel Modelling of Health Statistics—Leyland AH, Goldstein H, eds. (2001) Chichester: John Wiley & Sons. 159–74.
  12. Spiegelhalter DJ, Best NG, Carlin BP, van der Linde A. Bayesian measures of model complexity and fit. J Royal Statistical Soc B (2002) 64:583–639.[CrossRef]
  13. Camenzind E, Steg PG, Wijns W. Stent thrombosis late after implantation of first-generation drug-eluting stents. A cause for concern. Circulation (2007) 115:1440–55.[Free Full Text]
  14. Pfisterer M, Brunner-La Rocca HP, Buser PT, et al. Late clinical events after clopidogrel discontinuation may limit the benefit of drug-eluting stents: an observational study of drug-eluting versus bare-metal stents. J Am Coll Cardiol (2006) 48:2584–91.[Abstract/Free Full Text]
  15. Lagerqvist B, James SK, Stenestrand U, et al. Long-term outcomes with drug-eluting stents versus bare-metal stents in Sweden. N Engl J Med (2007) 356:1009–19.[Abstract/Free Full Text]
  16. Daemen J, Wenaweser P, Tsuchida K, et al. Early and late coronary stent thrombosis of sirolimus-eluting and paclitaxel-eluting stents in routine clinical practice: data from a large two-institutional cohort study. Lancet (2007) 369:667–78.[CrossRef][Medline]
  17. Stone GW, Moses JW, Ellis SG, et al. Safety and efficacy of sirolimus- and paclitaxel-eluting coronary stents. N Engl J Med (2007) 356:998–1008.[Abstract/Free Full Text]
  18. Spaulding C, Daemen J, Boersma E, et al. A pooled analysis of data comparing sirolimus-eluting stents with bare-metal stents. N Engl J Med (2007) 356:989–97.[Abstract/Free Full Text]
  19. Kastrati A, Mehilli J, Pache J, et al. Analysis of 14 trials comparing sirolimus-eluting stents with bare-metal stents. N Engl J Med (2007) 356:1030–9.[Abstract/Free Full Text]
  20. Beohar N, Davidson CJM, Kip KE, et al. Outcomes and complications associated with off-label and untested use of drug-eluting stents. JAMA (2007) 297:1992–2000.[Abstract/Free Full Text]
  21. Win HKM, Caldera AE, Maresh K, et al. Clinical outcomes and stent thrombosis following off-label use of drug-eluting stents. JAMA (2007) 297:2001–9.[Abstract/Free Full Text]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
30/2/186    most recent
fdn016v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Austin, D.
Right arrow Articles by Pell, J. P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Austin, D.
Right arrow Articles by Pell, J. P.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?