Is ophthalmology evidence based? A clinical audit of the emergency unit of a regional eye hospital

Comparing code-free and bespoke deep learning approaches in ophthalmology

AIM

Code-free deep learning (CFDL) allows clinicians without coding expertise to build high-quality artificial intelligence (AI) models without writing code. In this review, we comprehensively review the advantages that CFDL offers over bespoke expert-designed deep learning (DL). As exemplars, we use the following tasks: (1) diabetic retinopathy screening, (2) retinal multi-disease classification, (3) surgical video classification, (4) oculomics and (5) resource management.

METHODS

We performed a search for studies reporting CFDL applications in ophthalmology in MEDLINE (through PubMed) from inception to June 25, 2023, using the keywords 'autoML' AND 'ophthalmology'. After identifying 5 CFDL studies looking at our target tasks, we performed a subsequent search to find corresponding bespoke DL studies focused on the same tasks. Only English-written articles with full text available were included. Reviews, editorials, protocols and case reports or case series were excluded. We identified ten relevant studies for this review.

RESULTS

Overall, studies were optimistic towards CFDL's advantages over bespoke DL in the five ophthalmological tasks. However, much of such discussions were identified to be mono-dimensional and had wide applicability gaps. High-quality assessment of better CFDL applicability over bespoke DL warrants a context-specific, weighted assessment of clinician intent, patient acceptance and cost-effectiveness. We conclude that CFDL and bespoke DL are unique in their own assets and are irreplaceable with each other. Their benefits are differentially valued on a case-to-case basis. Future studies are warranted to perform a multidimensional analysis of both techniques and to improve limitations of suboptimal dataset quality, poor applicability implications and non-regulated study designs.

CONCLUSION

For clinicians without DL expertise and easy access to AI experts, CFDL allows the prototyping of novel clinical AI systems. CFDL models concert with bespoke models, depending on the task at hand. A multidimensional, weighted evaluation of the factors involved in the implementation of those models for a designated task is warranted.

An evidence-based medicine audit of the ophthalmic emergency services unit of King Abdulaziz University Hospital (KAUH)

PURPOSE

To improve the interventions provided for patients presenting with acute ophthalmic conditions to the ophthalmic emergency unit through applying the best available evidences from quality literature for managing such conditions.

METHODS

A retrospective cohort study at a tertiary eye care university hospital in Riyadh, Saudi Arabia, involving a two-phase audit of diagnosis-intervention was conducted. The first phase was done retrospectively for the duration from April 1 to May 30, 2014, after disseminating the results of the first phase to King Abdulaziz University Hospital ophthalmology department staff, and the second phase was done retrospectively for the duration from November 1 to December 30, 2015. The validity of outcomes was assessed through a literature search using Medline and the Cochrane Database of Systematic Reviews. The participants were masked on the study objectives to avoid Hawthorne's phenomenon (prescribing bias).

RESULTS

In the first part of the audit, 73.2% out of 355 interventions were found to be evidence based. There was notable improvement of 80.9% in the number of evidence-based interventions in the second part of the audit. This improvement was statistically significant (p = 0.017).

CONCLUSION

Evidence-based medicine audit can be a helpful tool to assess the performance and can lead to quality improve of the provided care by reducing the number of medical errors and refining medical decisions and interventions.

Patching for corneal abrasion

BACKGROUND

Published audits have demonstrated that corneal abrasions are a common presenting eye complaint. Eye patches are often recommended for treating corneal abrasions despite the lack of evidence for their use. This systematic review was conducted to determine the effects of the eye patch when used to treat corneal abrasions.

OBJECTIVES

The objective of this review was to assess the effects of patching for corneal abrasion on healing and pain relief.

SEARCH METHODS

We searched CENTRAL (which contains the Cochrane Eyes and Vision Trials Register) (2016, Issue 4), Ovid MEDLINE, Ovid MEDLINE In-Process and Other Non-Indexed Citations, Ovid MEDLINE Daily, Ovid OLDMEDLINE (January 1946 to May 2016), EMBASE (January 1980 to May 2016), Latin American and Caribbean Health Sciences Literature Database (LILACS) (January 1982 to May 2016), System for Information on Grey Literature in Europe (OpenGrey) (January 1995 to May 2016), the ISRCTN registry (www.isrctn.com/editAdvancedSearch), ClinicalTrials.gov (www.clinicaltrials.gov) and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en). We did not use any date or language restrictions in the electronic searches for trials. We last searched the electronic databases on 9 May 2016. We also searched the reference lists of included studies, unpublished 'grey' literature and conference proceedings and contacted pharmaceutical companies for details of unpublished trials.

SELECTION CRITERIA

We included randomised and quasi-randomised controlled trials that compared patching the eye with no patching to treat simple corneal abrasions.

DATA COLLECTION AND ANALYSIS

Two authors independently assessed the risk of bias and extracted data. Investigators were contacted for further information regarding the quality of trials. The primary outcome was healing at 24, 48 and 72 hours while secondary outcomes included measures of pain, quality of life and adverse effects. We graded the certainty of the evidence using GRADE.

MAIN RESULTS

We included 12 trials which randomised a total of 1080 participants in the review. Four trials were conducted in the United Kingdom, another four in the United States of America, two in Canada, one in Brazil and one in Switzerland. Seven trials were at high risk of bias in one or more domains and one trial was judged to be low risk of bias in all domains. The rest were a combination of low risk or unclear.People receiving a patch may be less likely to have a healed corneal abrasion after 24 hours compared to those not receiving a patch (risk ratio (RR) 0.89, 95% confidence interval (CI) 0.79 to 1.00, 7 trials, 531 participants, low certainty evidence). Similar numbers of people in the patch and no-patch groups were healed by 48 hours (RR 0.97, 95% CI 0.91 to 1.02, 6 trials, 497 participants, moderate certainty evidence) and 72 hours (RR 1.01, 95% CI 0.97 to 1.05, 4 trials, 430 participants, moderate certainty evidence). Participants receiving a patch took slightly longer to heal but the difference was small and probably unimportant (mean difference (MD) 0.14 days longer, 95% CI 0 to 0.27 days longer, 6 trials, 642 participants, moderate certainty evidence).Ten trials reported pain scores. Most studies reported pain on a visual analogue scale (VAS). It was not possible to pool the data because it was skewed. In general, similar pain ratings were seen between patch and no-patch groups. Data from two trials reporting presence or absence of pain at 24 hours was inconclusive. There was a higher risk of reported pain in the patch group but wide confidence intervals compatible with higher or lower risk of pain (RR 1.51, 95% CI 0.86 to 2.65, 2 trials, 193 participants, low certainty evidence). Five trials compared analgesic use between the patch and no-patch groups. Data from three of these trials could be combined and suggested similar analgesic use in the patch and no-patch groups but with some uncertainty (RR 0.95, 95% CI 0.69 to 1.32, 256 participants, low certainty evidence). Frequently reported symptoms included photophobia, lacrimation, foreign body sensation and blurred vision but there was little evidence to suggest any difference in these symptoms in people with or without a patch.Activities of daily living (ADL) were assessed in one study involving children. There was little difference in ADL with the exception of walking which was reported to be more difficult with a patch on: VAS 1.7 cm (SD 2.1) versus 0.3 cm (SD 0.7).Complication rates were low across studies and there is uncertainty about the relative effects of patching or not patching with respect to these (RR 3.24, 95% CI 0.87 to 12.05, 8 trials, 660 participants, low certainty evidence). Three trials reporting rates of compliance to treatment found that 22% of participants did not have their eye patches during follow-up. No-patch groups generally received more adjuvant treatment with antibiotics or cycloplegics, or both, than the patch group. There were limited data on the effect of patching on abrasions greater than 10mm(2) in size.

AUTHORS' CONCLUSIONS

Trials included in this review suggest that treating simple corneal abrasions with a patch may not improve healing or reduce pain. It must be noted that, in these trials, participants who did not receive a patch were more likely to receive additional treatment, for example with antibiotics. Overall we judged the certainty of evidence to be moderate to low. Further research should focus on designing and implementing better quality trials and examining the effectiveness of patching for large abrasions.

A series of 3 cases of corneal abrasion with multiple etiologies

BACKGROUND

Most corneal abrasions are caused by mechanical injuries affecting the superficial epithelial layer. Although one of the functions of the eyelid is to protect the eye, its reaction time of about 425 msec is slower than many ocular insults; thus, corneal abrasions are among the most commonly occurring eye emergencies.

CASE REPORTS

Three cases of corneal abrasions with different etiologies are presented. The first case was a large abrasion of the cornea near the visual axis caused by a wood chip. A metal foreign body with rust was lodged in the cornea from metal grinding in the second case. The third case was iatrongenically induced by an A-scan probe while a fellow student was measuring the axial length of the eye.

CONCLUSION

Corneal abrasions are one of the most common ocular conditions presented to eye clinics or emergency departments. Although there are different etiologies of abraded corneal epithelium, current clinical management for most corneal abrasions involves a bandage contact lens, use of topical antibiotics, and cycloplegics. Large and central corneal abrasions, however, warrant a consultation with a corneal specialist.

Quality of reporting of key methodological items of randomized controlled trials in clinical ophthalmic journals

PURPOSE

To evaluate the reporting quality of key methodological items in randomized controlled trials (RCTs) in four general clinical ophthalmology journals.

METHODS

The reporting of 11 key methodological items in RCTs published in American Journal of Ophthalmology, Archives of Ophthalmology, British Journal of Ophthalmology and Ophthalmology in the year 2005 was assessed.

RESULTS

Sixty-seven eligible RCTs were assessed and the mean number of items reported was 6.3 per RCT. No significant difference in the mean number of items reported was found between the four journals (P=0.20). The most frequently reported item was ethics approval and informed consent (97.0%), followed by masking status (85.1%), description of withdrawals (76.1%), adverse events (73.1%), and intention-to-treat analysis (71.6%). Details on sequence generation, randomization restriction, allocation concealment, allocation implementation, patient flow diagrams, and sample size calculation were reported in <50% of the RCTs assessed. Both sample size and page length of the RCTs correlated with the number of methodological items reported (P=0.024 and P=0.008, respectively).

CONCLUSIONS

Similar to other specialties, rooms for improvement exist in the reporting of key methodological items of RCTs in clinical ophthalmic journals. Stricter adoption of the CONSORT statement might enhance the reporting quality of RCTs in ophthalmic journals.

Evidence-Based Practice in Oral and Maxillofacial Surgery: Audit of 1 Training Center

PURPOSE

To evaluate the proportion of evidence-based interventions in the field of oral and maxillofacial surgery in a regional training center.

PATIENTS AND METHODS

A prospective clinical audit was carried out within the discipline of Oral and Maxillofacial Surgery, University of Hong Kong in February 2004 for a period of 6 months to investigate the extent of evidence-based practice. Consecutive diagnosis and intervention pairs were identified and recorded through standardized charts in randomly selected clinical sessions. A corresponding literature search using Medline and the Cochrane Library was performed to identify best current evidence. Each pair was then analyzed and graded according to the best current evidence.

RESULTS

Of 500 cases, 273 were eligible for evaluation while the rest were excluded based on 4 defined exclusion criteria. A majority of interventions (n = 195, 71.4%) were found to be evidence-based. Seventy-eight (28.6%) interventions were found to be not evidence-based. Among the evidence, a majority (56.1%) was level 5 evidence, which are case series or systematic review/meta-analysis of case series, and 36% were level 3 or above, which are randomized control trial (RCT) (level 3), meta-analysis of RCTs (level 2), or systematic review of RCTs (level 1). There was no statistically significant difference in the proportion of evidence-based practice between specialists and trainees in oral and maxillofacial surgery who saw and treated patients.

CONCLUSION

This study demonstrated that most interventions prescribed in this oral and maxillofacial surgery training center were evidence-based, and the proportion was comparable with that reported by other specialties.

An audit of corneal abrasion management following the introduction of local guidelines in an accident and emergency department

BACKGROUND AND OBJECTIVES

Corneal abrasions are a common presentation to accident and emergency (A&E) departments. Patients can be treated and discharged by A&E staff without the need for an ophthalmologist's attention; complicated cases, however, should be recognised and referred. Inexperience and limited training in ophthalmology may lead to suboptimal patient care and inappropriate use of ophthalmology outpatient clinics. Issues of poor documentation may also arise. The purpose of this audit was to assess the effect of guidelines on the management of corneal abrasion by A&E staff.

METHODS

A retrospective case note audit was performed to assess current management of corneal abrasions. Guidelines for management of corneal abrasions were formulated following a literature search and collaboration between A&E and ophthalmology staff. A prospective case note audit was undertaken to assess management after introduction of the guidelines.

RESULTS

A total of 51 cases were audited before the introduction of the guidelines and 57 cases after. Following the introduction of the guidelines documentation of visual acuity increased to 93% and specific enquiry into contact lens wear rose from 35.3% to 71.9%. A&E staff stopped giving out local anaesthetic eye drops. The follow up profile also improved; appropriate patient discharges increased by 40% whereas inappropriate referrals to eye clinic dropped by 75%. More patients were given abrasion advice (a 101% increase).

CONCLUSIONS

A&E staff members are capable of managing corneal abrasions if they are given guidance and some training. This audit identified shortcomings in current management and showed that guidelines can significantly improve clinical practice.

Evidence-based practice in acute ophthalmology

AIM

To study the overall level of evidence-based practice in an accident and emergency eye unit in the UK and evaluate the extent of evidence-based practice by ophthalmologists and nurse practitioners (NPs).

METHODS

The case notes of all new patients attending our accident and emergency department for a period of 1 week were reviewed prospectively. For each case, the diagnosis at presentation and its intervention was ascertained. A literature search using Medline and Cochrane library was performed to assess the level of evidence provided for each diagnosis-intervention pair. Each diagnosis-intervention pair was evaluated and graded to assess the best level of evidence provided.

RESULTS

A total of 474 new patients attended the eye casualty during the study period. No diagnosis was made in 27 cases and they were excluded from the analysis. Out of 447 patients, 308 (68.9%) cases were managed by NPs and 139 (31.1%) by doctors. A total of 294 (65.8%) patient interventions were based on evidence from systematic reviews, meta-analysis and randomised-controlled trials, of which 223 were managed by NPs and 71 by ophthalmologists. This difference was statistically significant with P<0.001. One hundred and seven (23.9%) interventions were supported by evidence from prospective and retrospective trials.

CONCLUSION

This study demonstrated that two-thirds of interventions were based on the highest level of evidence and was comparable to studies performed in other specialities. It also highlights the advanced role of trained NPs in acute ophthalmology.

Patching for corneal abrasion

BACKGROUND

Recent audits show that corneal abrasion is a common presenting eye complaint. Eye patches are often recommended for treating corneal abrasions despite the lack of evidence for their use. This systematic review was conducted to determine the effects of the eye patch when used to treat corneal abrasions.

OBJECTIVES

The objective of this review was to test the hypothesis that patching an eye following a corneal abrasion improves healing or provides pain relief.

SEARCH STRATEGY

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (which contains the Cochrane Eyes and Vision Group Trials Register) on The Cochrane Library (2005, Issue 2), MEDLINE (1966 to April 2005), EMBASE (1980 to April 2005), LILACS (13 April 2005), NRR (2005, Issue 2) and SIGLE (December 2004). There were no language or date restrictions in the searches. We also searched the reference lists of included studies, unpublished 'grey' literature and conference proceedings and contacted pharmaceutical companies for details of unpublished trials.

SELECTION CRITERIA

We included randomised and quasi-randomised controlled trials that compared patching the eye with no patching to treat simple corneal abrasions.

DATA COLLECTION AND ANALYSIS

Two authors independently assessed trial quality and extracted data. We contacted investigators for further information regarding quality of trials. The primary outcome was healing of the corneal epithelium and secondary outcomes were related to pain.

MAIN RESULTS

Eleven trials, which randomised a total of 1014 participants, were included in the review. Meta-analysis of seven studies with dichotomous healing outcomes favoured no patching on the first day of healing (risk ratio (RR) 0.89, 95% Confidence Interval (CI) 0.79 to 0.99). For days two and three there was no significant difference between the two groups. Of the nine trials that measured pain scores two favoured no patching and none favoured patching. Complication rates were low and no differences were noted in these between the two groups. No-patch groups generally received more adjuvant treatment with antibiotics and/or cycloplegics than the patch group which is an important confounding factor.

AUTHORS' CONCLUSIONS

Treating simple corneal abrasions with a patch does not improve healing rates on the first day post-injury and does not reduce pain. In addition, use of patches results in a loss of binocular vision. Therefore it is recommended that patches should not be used for simple corneal abrasions. Further research should focus on large (greater than 10 mm(2)) abrasions.

Is general inpatient obstetrics and gynaecology evidence-based? A survey of practice with critical review of methodological issues

BACKGROUND

To examine the rates of evidence-supported care provided in an obstetrics-gynaecology unit.

METHODS

The main diagnosis-intervention set was established for a sample of 325 consecutive inpatient admissions in 1998-99 in a prospective study in a UK tertiary care centre. A comprehensive literature search was conducted to obtain the evidence supporting the intervention categorised according to the following hierarchy: Grade A, care supported by evidence from randomised controlled trials; Grade B, care supported by evidence from controlled observational studies and convincing non-randomised evidence; and Grade C, care without substantial research evidence.

RESULTS

Of the 325 admissions, in 135 (42%) the quality of care was based on Grade A evidence, in 157 (48%) it was based on Grade B evidence, and in 33 (10%) it was based on Grade C evidence. The patterns of care were not different amongst patients sampled in 1998 and 1999.

CONCLUSION

A significant majority (90%) of obstetric and gynaecological care was found to be supported by substantial research evidence.

Gaps between best evidence and practice: causes for concern

Overseas studies that aim to quantify the evidence base of conventional medical care give varying estimates, but many of these studies have potential for bias. We do not know how much of the total healthcare Australians receive is based on the best available evidence; studies of a number of specific conditions show that there are gaps between what is known and what happens in practice. The National Institute of Clinical Studies aims to identify and test systemic approaches to embed ongoing review and uptake of evidence into routine clinical care.

Randomised controlled trial of clinical decision support tools to improve learning of evidence based medicine in medical students

OBJECTIVE

To assess the educational effectiveness on learning evidence based medicine of a handheld computer clinical decision support tool compared with a pocket card containing guidelines and a control.

DESIGN

Randomised controlled trial.

SETTING

University of Hong Kong, 2001.

PARTICIPANTS

169 fourth year medical students.

MAIN OUTCOME MEASURES

Factor and individual item scores from a validated questionnaire on five key self reported measures: personal application and current use of evidence based medicine; future use of evidence based medicine; use of evidence during and after clerking patients; frequency of discussing the role of evidence during teaching rounds; and self perceived confidence in clinical decision making.

RESULTS

The handheld computer improved participants' educational experience with evidence based medicine the most, with significant improvements in all outcome scores. More modest improvements were found with the pocket card, whereas the control group showed no appreciable changes in any of the key outcomes. No significant deterioration was observed in the improvements even after withdrawal of the handheld computer during an eight week washout period, suggesting at least short term sustainability of effects.

CONCLUSIONS

Rapid and convenient access to valid and relevant evidence on a portable computing device can improve learning in evidence based medicine, increase current and future use of evidence, and boost students' confidence in clinical decision making.