Developing decision support tools incorporating personalised predictions of likely visual benefit versus harm for cataract surgery: research programme
Sparrow JM., Grzeda M., Frost A., Liu C., Johnston RL., Scanlon P., Pithara C., Elliott D., Donovan J., Joseph-Williams N., Holland-Hart D., Donachie PHJ., Dixon P., Kandiyali R., Taylor H., Breheny K., Sterne J., Hollingworth W., Evans D., Fox F., Theodoropoulou S., Hughes R., Quinn M., Gray D., Benjamin L., Loose A., Edwards L., Craggs P., Paget F., Kapoor K., Searle J.
Background Surgery for established cataract is highly cost-effective and uncontroversial, yet uncertainty remains for individuals about when to proceed and when to delay surgery during the earlier stages of cataract. Objective We aimed to improve decision-making for cataract surgery through the development of evidence-based clinical tools that provide general information and personalised risk/benefit information. Design We used a mixed methodology consisting of four work packages. Work package 1 involved the development and psychometric validation of a brief, patient self-reported measure of visual difficulty from cataract and its relief from surgery, named Cataract Patient-Reported Outcome Measure, five items (Cat-PROM5). Work package 2 involved the review and refinement of risk models for adverse surgical events (posterior capsule rupture and visual acuity loss related to cataract surgery). Work package 3 involved the development of prediction models for the Cat-PROM5-based self-reported outcomes from a cohort study of 1500 patients; assessment of the validity of preference-based health economic indices for cataract surgery and the calibration of these to Cat-PROM5; assessment of patients’ and health-care professionals’ views on risk–benefit presentation formats, the perceived usefulness of Cat-PROM5, the value of personalised risk–benefit information, high-value information items and shared decision-making; development of cataract decision aid frequently asked questions, incorporation of personalised estimates of risks and benefits; and development of a cataract decision quality measure to assess the quality of decision-making. Work package 4 involved a mixed-methods feasibility study for a fully powered randomised controlled trial of the use of the cataract decision aid and a qualitative study of discordant or mismatching perceptions of outcome between patients and health-care professionals. Setting Four English NHS recruitment centres were involved: Bristol (lead centre), Brighton, Gloucestershire and Torbay. Multicentre NHS cataract surgery data were obtained from the National Ophthalmology Database. Participants Work package 1 – participants (n = 822) were from all four centres. Work package 2 – electronic medical record data were taken from the National Ophthalmology Database (final set > 1M operations). Work package 3 – cohort study participants were from Bristol (n = 1200) and Gloucestershire (n = 300); qualitative and development work was undertaken with patients and health-care professionals from all four centres. Work package 4 – Bristol, Brighton and Torbay participated in the recruitment of patients (n = 42) for the feasibility trial and recruitment of health-care professionals for the qualitative elements. Interventions For the feasibility trial, the intervention was the use of the cataract decision aid, incorporating frequently asked questions and personalised estimations of both adverse outcomes and self-reported benefit. Main outcome measures There was a range of quantitative and qualitative outcome measures: questionnaire psychometric performance metrics, risk indicators of adverse surgical events and visual outcome, predictors of self-reported outcome following cataract surgery, patient and health-care practitioner views, health economic calibration measures and randomised controlled trial feasibility measures. Data sources The data sources were patient self-reported questionnaire responses, study clinical data collection forms, recorded interviews with patients and health-care professionals, and anonymised National Ophthalmology Database data. Results Work package 1 – Cat-PROM5 was developed and validated with excellent to good psychometric properties (Rasch reliability 0.9, intraclass correlation repeatability 0.9, unidimensionality with residual eigenvalues ≤ 1.5) and excellent responsiveness to surgical intervention (Cohen delta –1.45). Work package 2 – earlier risk models for posterior capsule rupture and visual acuity loss were broadly affirmed (C-statistic for posterior capsule rupture 0.64; visual acuity loss 0.71). Work package 3 – the Cat-PROM5-based self-reported outcome regression models were derived based on 1181 participants with complete data (R 2 ≈ 30% for each). Of the four preference-based health economic indices assessed, two demonstrated reasonable performance. Cat-PROM5 was successfully calibrated to health economic indices; adjusted limited dependent variable mixture models offered good to excellent fit (root-mean-square error 0.10–0.16). The personalised quantitative risk information was generally perceived as beneficial. A cataract decision aid and cataract decision quality measure were successfully developed based on the views of patients and health-care professionals. Work package 4 – data completeness was good for the feasibility study primary and secondary variables both before and after intervention/surgery (data completeness range 100–88%). Considering ability to recruit, the sample size required, instrumentation and availability of necessary health economic data, a fully powered randomised controlled trial (patients, n = 800, effect size 0.2 standard deviations, power 80%; p = 0.05) of the cataract decision aid would be feasible following psychometric refinement of the primary outcome (the cataract decision quality measure). The cataract decision aid was generally well-received by patients and health-care professionals, with cautions raised regarding perceived time and workload barriers. Discordant outcomes mostly related to patient dissatisfaction, with no clinical problem found. Limitations The National Ophthalmology Database data are expected to include some errors (mitigated by large multicentre data aggregations). The feasibility randomised controlled trial primary outcome (the cataract decision quality measure) displayed psychometric imperfections requiring refinement. The clinical occurrence of discordant outcomes is uncommon and the study team experienced difficulty identifying patients in this situation. Future work Future work could include regular review of the risk models for adverse outcomes to ensure currency, and the technical precision of complex-numbers analysis of refractive outcome to invite opportunities to improve post-operative spectacle-free vision. In addition, a fully powered randomised controlled trial of the cataract decision aid would be feasible, following psychometric refinement of the primary outcome (the cataract decision quality measure); this would clarify its potential role in routine service delivery. Conclusions In this research programme, evidence-based clinical tools have been successfully developed to improve pre-operative decision-making in cataract surgery. These include a psychometrically robust, patient-reported outcome measure (Cat-PROM5); prediction models for patient self-reported outcomes using Cat-PROM5; prediction models for clinically adverse surgical events and adverse visual acuity outcomes; and a cataract decision aid with relevant general information and personalised risk/benefit predictions. In addition, the successful mapping of Cat-PROM5 to existing health economic indices was achieved and the performances of indices were assessed in patients undergoing cataract surgery. A future full-powered randomised controlled trial of the cataract decision aid would be feasible (patients, n = 800, effect size 0.2 standard deviations, power 80%; p = 0.05). Trial registration This trial is registered as ISRCTN11309852. Funding This project was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme and will be published in full in Programme Grants for Applied Research; Vol. 10, No. 9. See the NIHR Journals Library website for further project information.