Glaucoma has been estimated to account for 13% of new1 and 23% of follow-up appointments to hospital outpatient clinics.2 In addition, current trends suggest that improvements in detection, more aggressive treatment regimes and an ageing population will increase the demand for these services.3

Since the introduction of the concept of patient-centred care and patient choice in the NHS plan (2000),4 much emphasis has been placed on designing the provision of NHS care around the expectations of patients. One of the principal themes is the provision of some of the care that patents receive in the community. This ambition has recently been further re-enforced by the publication of the Darzi report.5

In ophthalmology, traditionally medical ophthalmic services have been provided by hospital eye service (HES)-based consultant led teams. For glaucoma, recent trends have seen other care providers undertaking this role such as ophthalmic opticians,3 6 specialist nurses3 and potentially general practitioners with a special interest (GPSi). Plurality of potential service providers for some aspects of ophthalmic care therefore now exists.

With all of the proposed changes in service delivery, we wanted to obtain a patients’ perspective as to what they wished for from follow-up of their glaucoma. The aim of this study was therefore to elucidate what factors are important to glaucoma patients when choosing a provider for their glaucoma follow-up review. This information would greatly help when planning and providing a glaucoma service and offer us an insight into aspects of service provision considered important by patients.

One way of ascertaining a patients’ preference for a product or service is by assessing “trade off” using conjoint analysis (CA). Briefly, conjoint analysis is a technique that has recently been used in healthcare research particularly to determine patients preferences for service development7 8 and in ophthalmology to determine patients’ preferences for treatment of various ophthalmic conditions.9 10 In the full concept (also known as full-profile) method, the respondent is asked to rank a set of profiles according to preference. All the factors (attributes) of interest are represented at different levels on each profile so that a complete description (or full concept) is provided.

The full concept method of analysis therefore generates a number of profiles or scenarios with the attributes of interest for the service package or product represented, each attribute represented by a number of options (levels) to describe that attributes’ possible characteristics. It allows the relative importance of different attributes to be assessed and shows what features individuals are prepared to trade to obtain what they think is most important. Utility scores are generated for each of the attribute levels and can be used to find the relative importance of each attribute level. It is therefore a quantitative technique that requires the subjects to make decisions and rank a series of scenarios.

METHODS

Ethic committee approval was granted by the Nottingham Ethics Committee; all patients were supplied with an information sheet and signed a consent form prior to participation in the study.

A pilot study was performed to determine which factors were important with respect to outpatient services for glaucoma patients. A list of potential attributes which were thought to be important was compiled by the ophthalmic team. Ten patients were interviewed and were asked about the importance of these factors with regard to the delivery of an outpatient service. In addition, they were invited to highlight any other factors they felt were important. Following these interviews, five factors were identified as most important by patients, and these factors formed the basis for the conjoint analysis (CA) scenarios. These factors were:

• travel time to clinic (three levels);

• training of healthcare professional (two levels);

• number of visits required to complete the episode (two levels);

• ease of access (two levels);

• clinic waiting time (two levels).

We chose as many levels as possible without causing an unacceptable increase in the number of cards that had to be ranked. Travel time to clinic had three levels because we thought from our pilot study that it would be an important factor, and we wished to capture it in as much detail as possible

In order to ensure that the glaucoma patients undertaking the study had a uniform understanding of the different attributes used to create the CA scenarios, definitions for each of the attributes were developed (box 1) to inform the patients in their decision-making and ensure consistency in interpretation of the various scenarios presented.

With these factors and factor levels, it is possible to generate 48 different scenarios (3×2×2×2×2 = 48), and rather than ask respondents to rank all 48 scenarios, we used a factor design (termed “orthoplan” in SPSS)11 to randomly generate an “orthogonal array” of eight scenarios. In performing this, it is assumed that interactions between the different factors are negligible. An example of one scenario is shown in box 2. In addition, two holdout scenarios were generated. These were generated from another random plan and not the experimental orthogonal plan in order to assess validity of the estimated utilities.

Interviews

One hundred patients attending clinic when interviewers were present agreed to participate; seven patients declined to be interviewed. Three patients were randomly selected from the clinic list on days that interviewers were present. Two physicians, both ophthalmologists, conducted the interviews (JB & AB). All the patients were required to be attending for a glaucoma review and be willing to participate in the interview process. The interview had three stages:

1. collection of demographic and visual function data;

2. completion of the VFQ-25 questionnaire of visual function;12

3. ranking of the 10 scenario cards.

The clinical data recorded were the mean defects from their Humphrey visual fields (24-2 SITA standard program) both at presentation and at the time of interview (classified as “best eye” and “worst eye” visual-field defects), the number of topical medications being used, the total number of trabeculectomies undertaken and the presenting and the current intraocular pressures and best eye and worst eye logMAR visual acuity.

The demographic data collected were age, gender, postcode, distance patient lived from hospital and the age that they left full-time education.

Social deprivation was calculated from the postcode using the Index of Multiple Deprivation (IMD) score database for social deprivation. IMD 2004 is a composite deprivation index containing seven domains: income, employment, health and disability, education, housing, environment and crime. The IMD 200413 is published at Super Output Area (SOA) levels for the whole of England in a downloadable table14

The distance of the patients’ home from the current clinic at Queens Medical Centre was calculated by using the postcode references for the hospital and the patients’ postcode using the “shortest route” option.

Data entry and analysis

The demographic, visual-function data and ranking data were entered into an Excel spreadsheet and then transferred to SPSS. The conjoint analysis was performed using the “conjoint” procedure in SPSS categories. This procedure takes the ranking of the different scenarios for each participant and, through a series of linear regressions, generates utility scores for each factor level. Each factor was specified as linear. SPSS calculates a regression coefficient for each factor, and the utility scores are the product of the coefficients times the factor level. The relative importance of each factor can also be expressed in percentage terms. This is done by taking the range of utility scores for any factor (highest minus lowest), dividing it by the sum of all the utility ranges and multiplying by 100.

Then, a multivariate linear regression (p for entry <0.05 and for removal >0.1) was undertaken. Variables entered in the regression model were minimum and maximum intraocular pressure (IOP) at presentation, current maximum and minimum eye IOP, minimum and maximum Humphrey visual-field (HVF) mean defect at presentation, minimum and maximum current HVF mean defect, previous number of trabeculectomies, current number of glaucoma drop medications being used, maximum and minimum logMAR visual acuity scores currently and at presentation, VFQ 25 score, age, gender, race, age left education distance from hospital, IMD score and number of years with glaucoma (table 1).

RESULTS

One hundred patients were interviewed, and four were excluded: one due to incomplete data and three due to the number and nature of “reversals,” which will be discussed below. The study population analysed therefore consisted of 47 females and 49 males, aged between 33 and 88 (mean age 67.5 years, median 69). Eighty-eight were Caucasian and eight Afro-Caribbean. The average age at which the group left education was 15.7 years. They had had glaucoma for an average for 5.5 years (ranging from 1 to 24 years, median 3 years) (table 1). Twelve patients had had glaucoma surgery, nine had had one trabeculectomy, two had had two, and one had had three.

The initial analysis showed a large number of reversals (seven subjects had three reversals, 23 had two reversals, and 19 had one reversal). A reversal is when a patient appears to prefer something that is perceived as the poorer choice such as they choose to travel further or prefer two visits rather than one. When this number is high, there is concern that the task had not been correctly understood. Inspection showed that the great majority of the reversals took place for factors of least importance. For three cases, however, the reversals involved the factors of greatest importance, and this suggests that these three subjects had not correctly understood the task. They were therefore eliminated from further analyses, and the conjoint analyses were repeated. The analysis presented therefore is on a total of 96 patients. The outcomes of conjoint analysis are summarised in table 2.

Internal validity checks showed that the observed rankings for the eight scenarios and the utility scores from the conjoint analysis were highly correlated (Pearson R = 0.997; p = 0.0001) and the two holdouts showed good correlation with the ranking predicted by the conjoint analysis (Kendall tau = 1.0). Multivariate linear regression (p for entry <0.05 and for removal >0.1) showed there were no significant predictors among the independent variables for the utility scores of the two most important factors, the level of training and the travel time. There were also no predictors for the ease of access or the waiting time utility scores.

There were, however, two factors that predicted the utility score for the number of visits. The first was distance from clinic with patients who lived further away preferring a one-stop service. The second was the visual-field defect in the better of the two eyes (this is an indicator of severity of disease). Those with a larger (more severe) defect also preferred the one-stop service (table 3).

DISCUSSION

This is the first study to our knowledge that has addressed patients’ preferences in a systematic way for follow-up care of treated glaucoma. One previous study has demonstrated patient satisfaction with optometrist-led care provision in the community,15 but not in the context of a package of care as presented here. In that study, only a small group of patients deemed suitable for care in the community were included, whereas all patients attending the glaucoma clinic were deemed eligible for this study, thus providing a broader perspective of patient opinion.

This study clearly identifies that for patients the most important of the factors presented to them was the travel time to the location at which their appointment was provided (35.4% importance), and within this factor the level with the highest utility score for desirability was <30 min the lowest of the options offered. This clearly supports ambitions to bring outpatient services closer to the patient5 and supports observations from previous studies.16

The next most important attribute was the type of health professional delivering the service (25.9% importance). In this study, there was a clear patient preference for doctors to deliver this service. This could indicate a potential bias within this cohort, as all patients were already being looked after in a consultant-led clinic in the hospital, but this was also the case in the Bristol study, and patients still preferred optometrist run community-based follow-up.15 In addition, within our current clinics, specially trained ophthalmic opticians also see patients. However, patients’ knowledge of these opticians was not investigated in this study, and it is possible that despite the guidance provided, many patients may simply have perceived opticians as the high street optician they visit from time to time for spectacles. A further explanation is the nature of the disease severity in the current clinics; the majority of these patients were diagnosed as having some form of glaucoma and associated visual disability with only four patients (two ocular hypertension and two glaucoma suspect) fulfilling the traditional criteria of low-risk disease that would have been followed up by optometrists in previous studies. This patient profile represents the current efforts of our department to transfer low-risk patients to community or hospital-based optometrists for further follow-up.

These two attributes account for over 60% of the importance viewed from the patients’ perspective, and it could be argued on the basis of these findings that patients would ideally like their care to be delivered by a doctor in the community.

Although the number of visits was not considered highly important by the group as a whole (13.8% importance), two groups of patients with further analysis were identified as considering this as an important feature: those who lived furthest from the hospital and those with significant visual disability. These observations are not surprising, as those living furthest from the hospital will experience the most inconvenience in attending an appointment especially if using public transport. Those with more severe visual disability will find making a journey a more demanding task, in addition to which patients with more advanced disease may require more frequent visits to the hospital for monitoring of their condition.

It is interesting, however, that patients generally were not troubled by the need for additional visits (apart from those with significant visual disability or long distances from the hospital), long waiting times to be seen by the care provider or the ease of access. It is possible that they assumed if the care was more local all of these other issues would be less important or less likely to be an issue.

This study has not attempted any economic analysis of provision of the difference scenarios generated; indeed the precise figures to undertake these calculations are often hard to acquire, show wide variation17 and are often disputed.18 19 However, if care were to be moved into the community, there would be several potential consequences. The economy of scale that exists in large hospital departments would be lost, as would the ability to organise additional tests on the spot.20 The ability of a consultant to be available to oversee a clinic of many patients with a team of doctors may also be limited due to the availability of sufficient room for several doctors in smaller community-based facilities. If the service were provided by a non-consultant, it is possible that quality of service would be affected with potentially more missed diagnoses and more referrals for a second opinion,21 22 which again may affect the economic effectiveness of the model and indeed the patient’s convenience.

An assumption in this approach is that the factors are independent. This is based on these factors being clearly identified as separate features by patients from the pilot study. This is a potential limitation of the methodology. We were careful to frame the questions so to reduce the risk of interaction and provided an accompanying definition leaflet for subjects participating in the study to ensure clarity in their understanding of the different attributes and their associated levels.

In conclusion, this study suggests that for this cohort of glaucoma patients, the most important features of follow-up identified were distance from home to clinic location and person providing the care. On the basis of this study, patients would prefer to be seen close to home within the community by a doctor.