A multilevel regression of patient-reported outcome measures after varicose vein treatment in England

Abstract

Introduction

The relationship between patient and hospital characteristics and their influence on quality of life (QoL) variance following varicose vein treatment is little understood. Whilst Patient-reported outcome measures (PROMs) can record postoperative outcomes, the actual comparison of PROMs between hospitals can be misleading when the clustered nature of varicose vein care is overlooked. Multilevel models can accommodate hierarchical data and therefore can provide a more accurate reflection of the relationship between patients and hospitals when investigating postoperative outcomes.

Methods

A multilevel model of PROMs was developed to analyse the relationship of patient characteristics (gender, age), postoperative outcomes (complications, postoperative satisfaction, treatment success) and hospital type (operative volume and if private or NHS institution) with the change in Aberdeen Varicose Vein Score (AVVQ) six months after varicose vein treatment.

Results

Between April 2010 and July 2014, some 24,460 PROMs from 162 hospitals were analysed. Whilst the majority of variance in AVVQ improvement was due to patient factors, a small but statistically significant amount of variance was detected due to differences between hospitals. Multilevel regression revealed that females saw a greater improvement in AVVQ, as did those who reported greater levels of treatment success and satisfaction. Patient age, complications, intervention, readmission, hospital size and hospital type were not significantly associated with AVVQ improvement.

Conclusion

Although QoL is intrinsically tied to an individual, hospitals can provide a small but potentially important benefit in AVVQ improvement following vein treatment. A patient-centred approach is therefore recommended to optimise patient outcomes.

Keywords

Multilevel model PROMs patient-reported outcome measures varicose veins

Introduction

Varicose veins are a common disease and estimated to affect up to half of all adults.^1–4 Whilst compression of the afflicted limb may provide some symptomatic relief, most patients find that treatment of the underlying venous incompetence by surgery or ablation results in better symptom resolution and enhanced quality of life.^5–7 Consequently, intervention for varicose veins is in high demand in the UK with at least 30,000 procedures performed annually.⁸

Since 2009, the Department of Health England has collected national statistics on the varicose vein service provided in England. A two-part questionnaire, the Patient Reported Outcome Measures (PROMs), is offered to all patients undergoing elective varicose vein treatment.⁸ The PROMs measure several patient outcomes such as postoperative satisfaction, treatment success, complications and, most importantly, quality of life (QoL). There are two measures of QoL used in the PROMs; the EuroQol 5-Dimension (EQ5D)⁹ and the Aberdeen Varicose Vein Questionnaire (AVVQ).^10,11 The EQ5D is a common general QoL instrument while the AVVQ is a more specific venous disease QoL instrument. Both are valid and sensitive measurements of QoL in patients with venous disease.^7,12,13 While all PROMs outcomes can be considered valuable, the National Institute for Health and Care Excellence (NICE) regards QoL improvement as one of the most important measures of a treatment outcome.¹⁴

While a cursory analysis of the PROMs data provides an overview of venous outcomes across the country and may even allow for a superficial comparison of care between hospital units, this ignores that these patients had, by definition, “clustered” into hospital groups for their vein treatment. Populations served by hospitals tend to be more similar when compared to a population as a whole and by not taking this into account, inaccurate conclusions can be drawn. By performing a multilevel linear regression, the stratification of the patient and hospital units can be controlled and the significance of certain patient and hospital characteristics correctly quantified.¹⁵ The aim of this study was to investigate the effect of patient and hospital characteristics as measured by PROMs on improvement of patient QoL as determined by the AVVQ.

Methods

Between April 2010 and July 2014, patients undergoing elective primary varicose vein treatment (as coded as per the Appendix) were offered the opportunity to complete pre- and post-treatment PROMs under the auspices of the Health and Social Care Information Centre (HSCIC).¹⁶ The preoperative PROMs record baseline demographics (age and gender) and baseline QoL using the EQ5D and AVVQ. After six months, the postoperative PROMs record complications, the treatment success, patient satisfaction and repeat the QoL using EQ5D and AVVQ. The AVVQ consists of 13 questions which calculate the level of disease severity on a scale of 0–100. A score of 0 would indicate no QoL impairment due to venous disease and a score of 100 would indicate significant QoL impairment due to venous disease. However, AVVQ scores close to 100 are unusual and most patients with uncomplicated varicose veins typically scoring between 10 and 30, and even patients with venous ulcers normally scoring between 30 and 60.¹³ The AVVQ reacts well to QoL improvement after venous intervention and consequently is one of the most popular disease-specific QoL measures used in venous research.^17–19 The EQ5D is a general measure of QoL and calculates an individual’s health state by asking five three-level ordinal questions. An EQ5D score of 1 is classed as a perfect health state without any impairment of the domains which determine QoL. Health states of 0 would represent a zero health state (i.e. death) and below 0 would represent a health state worse than death. Post-operative complications were self-reported by patients and recorded any incidence of allergy or reaction to a drug, post-operative bleeding, wound problems, urine problems or any need for further treatment or readmission after their initial treatment over the subsequent six months. Success and satisfaction were self-reported by patients on a five-level ordinal scale with 1 representing the best outcome and 5 the worst outcome. Individuals were censored of age and gender when fewer than six records were returned from a single provider to reduce the potential risk of patient identification.

The HSCIC database also recorded the volume of varicose vein operations performed at each individual institution using the national Hospital Episode Statistic (HES) database.⁸ Hospital units were classed as “NHS” if it was a NHS institution or “Private” if it was run by a private healthcare concern.

Outcomes

The primary outcome of this study was to explore AVVQ change following varicose vein treatment using a multilevel linear regression model. By treating all units of analysis as independent observations, a conventional linear regression model could inaccurately calculate the standard errors of the regression coefficients and therefore overstate statistical significance. By allowing for residual components to vary at different levels, a multilevel model will avoid this pitfall and provide a more reliable statistical conclusion.

Statistics

Descriptive statistics were calculated for both the patient-level and hospital-level variables. Hypothesis testing was performed using paired and unpaired t-tests for normally distributed data and Mann–Whitney U and Wilcoxon Signed Rank tests for non-normally distributed data. Correlation was performed using Spearman’s Rank testing, and if significant was further explored using linear regression for continuous data. Multivariable regression was performed to explore PROMs variables prior to multilevel regression. Multilevel linear regression is elaborated as follows.

Multilevel regression model

A series of multilevel hierarchical models were built in sequence to investigate the variance in AVVQ change. Models were of two levels, whereby patients constitute the first (lower) level and hospitals constitute the second (higher) level. The analysis was completed in three stages with each iteration increasing the model complexity and therefore explanatory capacity. The first model (Model 1) is a simple variance component model which measures the amount of variance in AVVQ observed across the two levels. This indicates how much of the overall variance observed is due to first- (patient) or second-level (hospital) characteristics.

The second model (Model 2) introduced independent variables at the first (patient) level to examine how much of the variance is due to the patient characteristics as follows: age, gender, allergy or reaction to drug, postoperative bleeding, wound problems, urine problems, further treatment, readmission and also self-reported success and treatment satisfaction scores. The third model (Model 3) introduced independent variables at the second higher (hospital) level to determine how much variance can be explained by the hospital size (varicose vein procedures performed per annum) and the hospital designation as either an NHS or Private hospital. Preliminary analysis was first undertaken in SPSS (Chicago, Version 22) while MLwiN (University of Bristol, Version 2.30) was used for subsequent multilevel analysis.

Results

Between April 2010 and July 2014, some 24,460 patients treated by 162 UK hospitals had returned varicose vein PROMs to the HSCIC. Baseline measurements and follow-up results are detailed in Tables 1 and 2, respectively. Overall, varicose vein treatment was associated with a significant improvement in patient QoL. Median (i.q.r.) AVVQ significantly improved from 17.9 (12.3–25.4) to 9.5 (4.0–17.6) (P < 0.001) with a mean (s.d.) improvement of 8.1 (9.8). Median (i.q.r) EQ5D also improved from 0.796 (0.725–0.848) to 1.000 (0.760–1.000), a mean (s.d.) increase of 0.093 (0.216).

Table 1.

Baseline characteristics.

Variable		Frequency (%)
Gender	Female	10530 (68.8)
Gender	Male	4783 (31.2)
Age	20–29	156 (1.0)
	30–39	1815 (11.9)
	40–49	3670 (24.0)
	50–59	3888 (25.4)
	60–69	4159 (27.2)
	70–79	1545 (10.1)
	80–89	80 (0.5)
Hospital class	NHS	23349 (95.5)
	Private	1111 (4.5)
	Operative volume	91 (39–194)^a

Median (interquartile range).

EQ5D: EuroQol 5 dimension.

Table 2.

Follow-up outcomes.

Variable		Frequency (%)
Satisfaction	Excellent	3731 (21.8)
	Very good	6217 (36.4)
	Good	4506 (26.4)
	Fair	1953 (11.4)
	Bad	671 (3.9)
Treatment success	Much better	11772 (69.0)
	A little better	3466 (20.3)
	About the same	1249 (7.3)
	A little worse	399 (2.3)
	Much worse	185 (1.1)
Allergy problem	Yes	454 (2.8)
Allergy problem	No	15550 (97.2)
Bleeding problem	Yes	2306 (14.2)
Bleeding problem	No	13889 (85.8)
Wound problem	Yes	2389 (14.6)
Wound problem	No	13982 (85.4)
Urine problem	Yes	0 (0.0)
Urine problem	No	15770 (100.0)
Further treatment	Yes	2596 (15.3)
Further treatment	No	14367 (84.7)
Readmission	Yes	861 (5.0)
Readmission	No	16207 (95.0)

Mean (standard deviation).

EQ5D: EuroQol 5 dimension.

Gender

A significant difference in quality of life was noted between the genders. Females reported worse AVVQ compared to males at baseline [female: 18.6 (13.5–25.5) vs male: 15.6 (10.1–23.2) P < 0.001] and worse scores after treatment [female: 10.6 (5.0–17.8) vs male: 6.7 (2.4–14.3) P < 0.001]. However, both groups saw an improved AVVQ score after treatment (P < 0.001 Wilcoxon) and mean improvement in AVVQ score was similar for both genders [female: −7.8 (9.4) vs male: −8.0 (9.9) P = 0.344]. This disparity in QoL between the genders was confirmed in EQ5D scores. Both the preoperative [female: 0.796 (0.725–0.796) vs male: 0.796 (0.725–1.000) P < 0.001] and postoperative [female 0.850 (0.760–1.000) vs male: 1.000 (0.796–1.000) P < 0.001] EQ5D scores showed females at a disadvantage. However, again, there was no difference in actual improvement for either genders [female + 0.052 (0.000–0.204) vs male + 0.033 (0.000–0.204) P = 0.053].

Age

As shown in Figure 1, there was a significant difference in median AVVQ across the age groups preoperatively and postoperatively (Kruskal Wallis P < 0.001). Older age groups typically reported higher AVVQ scores when compared to younger age groups. Most age groups were not drastically dissimilar in AVVQ score from their neighbouring age group except for a significant step in AVVQ severity noted between the age groups 40–49 to 50–59 in both preoperative (P < 0.001) and postoperative (P = 0.015) AVVQ scores. Although older age groups typically saw greater falls in AVVQ compared to younger patients (P < 0.001), no significant step in AVVQ improvement was detected across the age groups.

Figure 1.

Preoperative and postoperative Aberdeen Varicose Vein Score (AVVQ) and patient age.

Private vs pubic healthcare sectors

Private patients had less severe AVVQ scores compared to NHS patients in both preoperative [private 16.15 (11.29–22.88) vs NHS 18.00 (12.34–25.54) P < 0.001] and postoperative scores [private 7.09 (2.43–14.36) vs 9.41 (4.01–17.12) P < 0.001]. However, both private and NHS patients saw a similar amount of improvement after vein treatment [private −8.52 (8.81) vs NHS −8.07 (9.89) P = 0.144]. As shown in Table 3, no difference in postoperative complications was reported between NHS and private healthcare institutions. However, private patients did appear to be more satisfied and reported higher levels of treatment success compared to those treated in NHS hospitals.

Table 3.

Complications.

Complication	NHS	Private	P ^a
Allergy or reaction to drug	2.8%	3.4%	0.367
Bleeding	14.2%	15.1%	0.495
Wound problem	14.7%	12.9%	0.188
Urine problem	0%	0%	–
Further treatment	15.3%	16.3%	0.422
Readmission	5.1%	3.7%	0.078
Satisfaction	2 (2–3)	2 (1–3)	<0.001^b
Success	1 (1–2)	1 (1)	<0.001^b

χ².

Mann–Whitney U.

Unit size and outcomes

The number of procedures performed at an institution correlated significantly with AVVQ change (r²= 0.041, P < 0.001). Exploration of this relationship using linear regression suggested that every additional 10 vein treatments worsen the expected AVVQ improvement by 0.02 points.

Multivariable regression model

A multivariable linear regression model is detailed in Table 4. Gender, satisfaction, treatment success and requirement for additional treatments were significantly associated with AVVQ improvement. In holding all other variables constant, females improved an additional 0.542 AVVQ points compared to males, those “Excellently” satisfied improved an additional 5.325 AVVQ points compared to those “Badly” satisfied, those with a “Much better” treatment success improved an additional 11.397 AVVQ points compared to those with a “Much worse” treatment success and those who avoided additional treatments improved 0.629 AVVQ points more than those who did require further treatments. Age, allergy or drug reactions, bleeding problems, wound problems, readmission, hospital class and hospital size were not associated with AVVQ change.

Table 4.

Multivariable linear regression.

		Estimate	S.E.	P
Intercept		6.038	1.856	0.001
Patient variable
Gender	Female	−0.542	0.219	0.013
Age	30–39	−0.299	1.069	0.780
	40–49	−0.326	1.050	0.756
	50–59	−0.462	1.050	0.660
	60–69	−0.619	1.051	0.556
	70–79	−0.864	1.079	0.423
	80–89	−1.475	1.769	0.404
Satisfaction	Excellent	−5.325	0.776	<0.001
	Very good	−4.271	0.755	<0.001
	Good	−3.225	0.734	<0.001
	Fair	−1.813	0.708	0.010
Success	Much better	−11.397	1.292	<0.001
	A little better	−8.515	1.280	<0.001
	About the same	−6.159	1.251	<0.001
	A little worse	−4.733	1.341	<0.001
Allergy or reaction to drug	No	0.686	0.822	0.404
Bleeding problem	No	0.478	0.342	0.162
Wound problem	No	0.325	0.363	0.370
Further treatment	No	−0.629	0.275	0.022
Readmission	No	0.045	0.488	0.926
Hospital characteristics
	NHS	−0.376	0.636	0.555
	Volume (operations per year)	0.0001	0.0004	0.900

Note: Reference category: gender (male); age (20–29); satisfaction (bad); success (Much worse); allergy problem (yes); post-operative bleeding problem (yes); post-operative wound problem (yes); further treatment required (yes); readmission (yes); hospital sector (private hospital).

Multilevel linear regression model

The hierarchical linear regression model is detailed in Table 5. Using a simple model without any independent variables (Model 1), a statistically significant amount of variance between hospitals was detected (σ²μ₀ = 2.598 P < 0.001). The proportion of variance which was explained at the hospital level was 2.76% with the majority (97.24%) of the observed variance owing to patient-level factors.

Table 5.

Model estimates for multilevel regression of AVVQ change.

		Model 1 Intercept only			Model 2 Patient variables			Model 3 Hospital variables
		Estimate	S.E.	P	Estimate	S.E.	P	Estimate	S.E.	P
Fixed Effects	Intercept	−7.986	0.083	<0.001	5.679	1.758	<0.001	5.973	1.920	<0.001
Patient variables^a
Gender	Female				−0.673	0.272	0.013	−0.586	0.276	0.034
Age	30–39				−0.521	1.067	0.626	−0.447	1.068	0.676
	40–49				−0.469	1.047	0.654	−0.433	1.049	0.680
	50–59				−0.568	1.048	0.588	−0.523	1.049	0.618
	60–69				−0.791	1.049	0.451	−0.723	1.051	0.492
	70–79				−1.164	1.077	0.280	−0.935	1.079	0.386
	80–89				−0.909	1.707	0.594	−1.165	1.771	0.511
Satisfaction	Excellent				−5.398	0.769	<0.001	−5.235	0.773	<0.001
	Very good				−4.346	0.747	<0.001	−4.201	0.752	<0.001
	Good				−3.229	0.726	<0.001	−3.156	0.730	<0.001
	Fair				−1.828	0.701	0.009	−1.737	0.705	0.007
Success	Much better				−11.020	1.274	<0.001	−11.320	1.287	<0.001
	A little better				−8.327	1.262	<0.001	−8.498	1.275	<0.001
	About the same				−5.916	1.233	<0.001	−6.169	1.246	<0.001
	A little worse				−4.447	1.322	0.001	−4.706	1.336	<0.001
Allergy or reaction to drug	No				0.642	0.826	0.437	0.723	0.828	0.191
Postoperative Bleeding	No				0.530	0.359	0.139	0.516	0.362	0.154
Postoperative wound problem	No				0.361	0.377	0.338	0.314	0.380	0.409
Further treatment	No				−0.551	0.313	0.079	−0.471	0.317	0.137
Readmission	No				−0.020	0.496	0.968	0.086	0.503	0.864
Hospital variable^a
Procedures performed	(Grand mean centred)							0.0002	0.0005	0.624
NHS	Yes							−0.480	0.758	0.526
Random effects	Hospital-level variance	2.598	0.349	<0.001	1.602	0.428	<0.001	1.622	0.432	<0.001
	Patient-level variance	94.300	0.893	<0.001	82.960	1.282	<0.001	82.690	1.288	<0.001
	Intra-hospital correlation	2.76%			1.89%			1.96%

Note: S.E.: standard error

Reference category: gender (male), age (20–29), satisfaction (bad), success (much worse), allergy or drug reaction (yes), post-operative bleeding (yes), post-operative wound infection (yes), further surgery (yes), readmission (yes), hospital sector (private).

Patient characteristics were then introduced into the model (Model 2). Patient characteristics, as measured by the PROMs, could explain 12.7% of the overall fall in model variance when compared to the first model. However, these characteristics were only able to explain 12.03% of the actual reduction in variance within the patient level itself, suggesting that 87.97% of the variance seen between patients was due to presently unobserved patient characteristics. A much greater fall in variance was seen between hospitals with a 38.34% reduction in variance seen at the second level, suggesting that patient characteristics are comparatively much better predictors of outcomes. Hospital characteristics were then introduced into the model (Model 3). Total model variance improved minimally (0.003%), with hospital-level variation increasing slightly and patient-level variation decreasing slightly. Using scaled log-likelihood ratio χ² test to compare models, Model 3 was a better model fit (Log difference = 1029.64, df = 2, P < 0.001). In this final model, differences between hospitals explained 1.96% of the variance in AVVQ outcomes with 98.04% of the variance explained by patient-level factors.

The final model suggests that, in holding all other variables constant, females have 0.586 more AVVQ point improvement over males, patients “Excellently” satisfied have 5.235 more AVVQ points improvement over those with “Bad” satisfaction, and those “Much better” success have 11.320 more AVVQ points improvement than those with “Much worse” success. Age, allergy or drug problem, bleeding problem, wound problem, additional treatments and readmission were not significant explanatory variables, and neither hospital sector or hospital size were significant characteristics in explaining the variance in AVVQ improvement between hospitals.

Discussion

Whilst the vast majority of variation in AVVQ improvement following varicose vein treatment does appear to reside in differences between individual patients, a small but statistically significant amount of variance in AVVQ improvement can be attributed to differences between hospitals. Most of the factors that affect AVVQ variance are, however, as of yet still unclear. Current PROMs suggest that gender, treatment success and satisfaction are patient characteristics which are strongly associated with AVVQ improvement. Reasons for the difference between hospitals are currently unknown and evidently such intra hospital variance observed is not explained by hospital size nor hospital type. It therefore appears that a significant proportion of unobserved patient and hospital characteristics remain to be identified which may potentially help explain this unaccounted variance. While the presence of “unknown unknowns” may cause some unease, it is a fortuitous advantage of multilevel models modelling that such gaps in knowledge can be clearly quantified and delineated several noteworthy results have been extracted from the data regardless.

Firstly, females are more represented within the PROMs and respond better to treatment. Although women are more likely to be referred for treatment compared to males, it is important to note that females suffer a significantly disproportional amount of venous disease QoL impairment compared to men, typically scoring 3–4 AVVQ points with the same clinical severity potentially explaining this phenomenon.^20,21 Whilst the burden of venous disease is worse for women, it also appears that females are generally more susceptible to developing venous disease over their lifetime and consequently this explains why more females undergo treatment and contribute a higher proportion of PROMs.^2,22 This study indicates that once the burden of venous disease is lifted, women respond with a much greater fall AVVQ, thereby confirming the additional impairment women experience due to venous disease on their quality of life.

Second, it is understandable that patients report higher levels of satisfaction and perceive better treatment outcomes if they find that their QoL has significantly improved after treatment. Despite most patients reporting reasonable to excellent treatment outcomes, it is a concern that so many patients report “fair” to “bad” satisfaction. Two decades ago, a study by Davies et al.²³ suggested that 26% of NHS and 13% of private varicose vein patients report high levels of dissatisfaction after vein surgery. Since conventional vein surgery has become a third choice after minimally invasive surgical techniques, it is hoped that patient care and outcomes may improve with more specialised and less traumatic endovenous treatments.²⁴ However, patients seem to be broadly satisfied with either conventional or endovenous surgery in the short term, and long-term satisfaction appears to be more tied into vein recurrence or return of symptoms.^19,25 Just as larger improvements in AVVQ are understood to relate to a higher perception of treatment success, satisfaction is also more likely to be increased in those with stronger AVVQ improvements. Improving reported satisfaction and perceived treatment success will therefore respond to techniques and methods which can enhance QoL improvements following vein treatment.

Third, the fact that most of the PROMs and hospital characteristics measured actually do not relate to QoL improvement is intriguing. Although older patients present for treatment with a typically worse AVVQ and appear to experience a greater benefit from treatment, age does not apparently influence the eventual improvement in AVVQ score once other patient factors are controlled. It is also counterintuitive that complications do not affect QoL improvement. The impact of complications on QoL appears to be either limited or short lived with little bearing on the final AVVQ score. Whilst problems such bleeding and allergy or reaction to drugs can be expected to be transient issues, more complicated issues such as wound problems, additional treatment and readmission do not appear to be a significant drag on typical QoL improvement. Indeed it is known that in the case of additional procedures, it appears that patients “catch up” QoL improvement once residual symptomatic veins are treated promptly, potentially explaining why additional procedures report similar QoL levels rather than signify worse outcomes.^26,27 It is also interesting to note that, although private healthcare patients seem to have higher levels of satisfaction and success, private care does not appear to translate into any additional improvement in QoL. Both private and NHS institutions appear to be equally adept at treating varicose veins, suggesting that the differences between hospitals are more likely to reside in general patient care practices common to all types of hospitals rather than activates which are distinct to either NHS or private hospitals.

There are several limitations of this study which should be noted. Patient characteristics recorded by the PROMs are quite limited and ideally should be expanded. The addition of more baseline and follow-up outcomes, such as disease-specific clinical severity measurements such as the CEAP²⁸ or VCSS²⁹ classification or pre- or postoperative duplex ultrasonography results, may help explain the observed variance but unfortunately such clinical measurements do not always relate to subsequent QoL improvements.^30–32 Operative and follow-up details are also missing. The impact of newer endovenous techniques could have been explored, as could various compression regimes and other postoperative management strategies. General issues related to the difficulty of completing a PROMs questionnaire, especially among older patients, should be noted. Censoring of data for facilities which perform few procedures may obscure the effect of small volume hospitals, especially as these smaller hospitals may preferentially treat less complex cases when compared to larger tertiary centres.

In conclusion, QoL improvement after varicose vein treatment is complicated and much remains to be understood. This large study appears to suggest that QoL improvement following varicose vein treatment depends mostly on the individual patient characteristics with only a small effect owing to the treatment centre where the procedure was performed. Regardless, varicose vein treatment is associated with significant improvements in QoL and methods of best practice need to be recognised and disseminated nationwide.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Appendix

References

Carpentier

Maricq

Biro

. Prevalence, risk factors, and clinical patterns of chronic venous disorders of lower limbs: a population-based study in France. J Vasc Surg 2004; 40: 650–659.

Beebe-Dimmer

Pfeifer

Engle

. The epidemiology of chronic venous insufficiency and varicose veins. Annal Epidemiol 2005; 15: 175–184.

Evans

Allan

Lee

. Prevalence of venous reflux in the general population on duplex scanning: the Edinburgh vein study. J Vasc Surg: official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 1998; 28: 767–776.

Maurins

Hoffmann

Losch

. Distribution and prevalence of reflux in the superficial and deep venous system in the general population – results from the Bonn Vein Study, Germany. Journal of vascular surgery: official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 2008; 48: 680–687.

Sell

Vikatmaa

Alback

. Compression therapy versus surgery in the treatment of patients with varicose veins: A RCT. Eur J Vasc Endovasc Surg 2014; 47: 670–677.

Michaels

Brazier

Campbell

. Randomized clinical trial comparing surgery with conservative treatment for uncomplicated varicose veins. Br J Surg 2006; 93: 175–181.

Michaels

Campbell

Brazier

. Randomised clinical trial, observational study and assessment of cost-effectiveness of the treatment of varicose veins (REACTIV trial). Health Technol Assess 2006; 10: 1–196, iii–iv.

HSCIC, http://www.hscic.gov.uk/hes (accessed 2014).

EuroQol

. EuroQol – a new facility for the measurement of health-related quality of life. Health Policy 1990; 16: 199–208.

10.

Garratt

Macdonald

Ruta

. Towards measurement of outcome for patients with varicose veins. Qual Health Care: QHC 1993; 2: 5–10.

11.

Smith

Garratt

Guest

. Evaluating and improving health-related quality of life in patients with varicose veins. J Vasc Surg: official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 1999; 30: 710–719.

12.

Kuet

Lane

Anwar

. Comparison of disease-specific quality of life tools in patients with chronic venous disease. Phlebology/Venous Forum R Soc Med 2014; 29: 648–653.

13.

Carradice

Mazari

Samuel

. Modelling the effect of venous disease on quality of life. Br J Surg 2011; 98: 1089–1098.

14.

Excellence National Institute for Health and Care Excellence. Methods for the development of NICE public health guidance. 3rd ed, http://wwwniceorguk/article/PMG4/chapter/1%20Introduction (2012, accessed December 2014).

15.

Snijders TAB and Bosker RJ. Multilevel analysis: an introduction to basic and advanced multilevel modeling. SAGE Publications, 2011.

16.

The Health and Social Care Information Centre, http://www.hscic.gov.uk/proms (2014, accessed October 2014).

17.

Brittenden

Cotton

Elders

. A randomized trial comparing treatments for varicose veins. New Engl J Med 2014; 371: 1218–1227.

18.

Rasmussen

Lawaetz

Bjoern

. Randomized clinical trial comparing endovenous laser ablation, radiofrequency ablation, foam sclerotherapy and surgical stripping for great saphenous varicose veins. Br J Surg 2011; 98: 1079–1087.

19.

Carradice

Mekako

Mazari

. Randomized clinical trial of endovenous laser ablation compared with conventional surgery for great saphenous varicose veins. Br J Surg 2011; 98: 501–510.

20.

Staniszewska

Tambyraja

Afolabi

. The Aberdeen varicose vein questionnaire, patient factors and referral for treatment. Eur J Vasc Endovasc Surg 2013; 46: 715–718.

21.

Conway

Nordon

Hinchliffe

. Patient-reported symptoms are independent of disease severity in patients with primary varicose veins. Vascular 2011; 19: 262–268.

22.

Bradbury

Evans

Allan

. What are the symptoms of varicose veins? Edinburgh vein study cross sectional population survey. BMJ 1999; 318: 353–356.

23.

Davies

Steffen

Cosgrove

. Varicose vein surgery: patient satisfaction. J R College Surg Edinburgh 1995; 40: 298–299.

24.

Marsden

Perry

Kelley

. Diagnosis and management of varicose veins in the legs: summary of NICE guidance. BMJ 2013; 347: f4279–f4279.

25.

Campbell

Vijay Kumar

Collin

. The outcome of varicose vein surgery at 10 years: clinical findings, symptoms and patient satisfaction. Annal R College Surg Engl 2003; 85: 52–57.

26.

Carradice

Mekako

Hatfield

. Randomized clinical trial of concomitant or sequential phlebectomy after endovenous laser therapy for varicose veins. Br J Surg 2009; 96: 369–375.

27.

Lane TR, Kelleher D, Shepherd AC, et al. Ambulatory varicosity avUlsion later or synchronized (AVULS): a randomized clinical trial. Ann Surg 2015; 261: 654–661.

28.

Eklof

Rutherford

Bergan

. Revision of the CEAP classification for chronic venous disorders: consensus statement. J Vasc Surg: official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 2004; 40: 1248–1252.

29.

Kakkos

Rivera

Matsagas

. Validation of the new venous severity scoring system in varicose vein surgery. J Vasc Surg: official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 2003; 38: 224–228.

30.

Lattimer

Kalodiki

Azzam

. The Aberdeen varicose vein questionnaire may be the preferred method of rationing patients for varicose vein surgery. Angiology 2014; 65: 205–209.

31.

Lattimer

Kalodiki

Azzam

. Responsiveness of individual questions from the venous clinical severity score and the Aberdeen varicose vein questionnaire. Phlebology/Venous Forum R Soc Med 2014; 29: 43–51.

32.

Shepherd

Gohel

Lim

. A study to compare disease-specific quality of life with clinical anatomical and hemodynamic assessments in patients with varicose veins. J Vasc Surg: official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 2011; 53: 374–382.