Telescreening for Retinopathy of Prematurity

Published on 08/03/2015 by admin

Filed under Opthalmology

Last modified 22/04/2025

Print this page

This article have been viewed 1010 times

Chapter 62 Telescreening for Retinopathy of Prematurity

Michael F. Chiang

Limitations of traditional care

Traditional screening for retinopathy of prematurity (ROP) involves indirect ophthalmoscopy at the bedside in the neonatal intensive care unit (NICU). Although this has been effective at identifying infants with severe treatment-requiring disease,¹^–³ there are important limitations. Ophthalmoscopic examinations are logistically difficult, require significant travel time and coordination, and are physiologically stressful for infants.⁴ Findings are documented using hand-drawn sketches, which are subjective and qualitative (Fig. 62.1). In addition, there may be variability in diagnosis of critical features such as zone I and plus disease, even among experts,^5,⁶ and there is enormous medicolegal liability. Surveys have found that the number of retinal specialists and pediatric ophthalmologists willing to manage ROP is decreasing for these reasons.⁷ Meanwhile, more infants are at risk for disease because of increasing premature birth rates and improved neonatal survival throughout the world.

Fig. 62.1 Documentation of traditional indirect ophthalmoscopy using annotated examination template. Limitations include subjective and qualitative documentation, and difficulty identifying change during serial examinations.

Telemedicine as an emerging approach

Telemedicine is an emerging approach with potential to improve the quality, delivery, and cost of care compared to traditional strategies. This may be particularly important in developing nations and other areas with limited accessibility to care. In this approach, clinical data and images are captured from the infant’s eyes by trained personnel in the NICU. Data are transmitted for review by a remote ophthalmologist, who sends management recommendations to the consulting physician. Examples of wide-angle images taken with a commercially available camera (RetCam; Clarity Medical Systems, Pleasanton, CA) are shown in Fig. 62.2. It has been shown that trained neonatal nurses can capture high-quality retinal images,^8,⁹ and that imaging may cause less physiological stress to infants than ophthalmoscopy with scleral depression.¹⁰

Fig. 62.2 Examples of ROP images captured by a trained neonatal nurse during routine screening using a wide-angle camera (RetCam; Clarity Medical Systems, Pleasanton, CA). Images demonstrate retinas with (A) no ROP, (B) type 2 ROP based on the presence of stage 3 disease in zone II, and (C) type 1 treatment-requiring ROP based on the presence of plus disease as well as posterior stage 3 disease.

In a telemedicine approach, the availability of archived retinal images would provide other advantages. Infant photographs could be directly compared to references such as the published standard photograph for plus disease,¹ and may be transmitted securely to experts for second opinions. Images provide objective documentation of clinical findings, improve recognition of disease progression, enhance communication, and create infrastructure for education and research.¹¹

Evaluation studies

Accuracy

Diagnostic accuracy of telemedicine for ROP has been evaluated since the late 1990s. Virtually all studies have used wide-angle digital images captured by a neonatal nurse, ophthalmologist, or ophthalmic photographer. Although these studies have varied in design and outcome measure, most have compared the diagnostic performance of telemedicine to a reference standard of dilated ophthalmoscopy. Schwartz et al. (19 eyes from 10 infants) examined the accuacy of telemedicine in a selected group of infants at 30–32 weeks postmenstrual age (PMA), all of whom had moderate or severe ROP. The sensitivity and specificity of telemedicine for detecting prethreshold or worse ROP were 89% and 100%, respectively.¹²

Since that time, research has involved larger and broader cohorts of consecutively enrolled infants. For diagnosis of any ROP regardless of severity, studies have demonstrated sensitivity of 46%–97% and specificity 49%–100% compared to a reference standard of indirect ophthalmoscopy ^8,^13,¹⁴ (Table 62.1). Generally, lower accuracy has been found while examining infants at lower PMA, and when detecting presence of mild ROP (e.g. peripheral stage 1). This is presumably because younger infants have milder disease with more subtle diagnostic features, and because it may be technically more difficult to image smaller eyes with increased media opacities.¹⁵

Table 62.1 Diagnostic accuracy of telescreening for detection of any ROP using images captured by wide-angle camera.* Reference standard was standard indirect ophthalmoscopy

Study	Outcome measures	Sensitivity/specificity
Roth et al., 2001¹⁴	Any ROP	0.82/0.94
Yen et al., 2002¹⁵	Any ROP at 32–34 wks PMA	0.46/1.00
Yen et al., 2002¹⁵	Any ROP at 38–40 wks PMA	0.76/1.00
Chiang et al., 2006¹³	Any ROP	0.82–0.86/0.49–0.96
Shah et al., 2006¹⁶	Any ROP	0.86/0.92
Chiang et al., 2007⁸	Any ROP at 31–33 wks PMA	0.73–0.94/0.89–0.97
Chiang et al., 2007⁸	Any ROP at 35–37 wks PMA	0.91–0.97/0.98–1.00
Dhaliwal et al., 2009¹⁷	Any ROP at 34 wks PMA or 4–6 wks CA	0.60/0.91

PMA, postmenstrual age; CA, chronological age.

^* RetCam; Clarity Medical Systems, Pleasanton, CA.

Other studies have examined accuracy of telemedicine for detecting clinically significant ROP (Table 62.2). Ells et al. (371 exams from 36 infants) found sensitivity 100% and specificity 96% for diagnosis of “referral-warranted ROP”* during serial examinations throughout an infant’s hospital course.¹⁸ Wu et al. (serial exams from 43 infants) performed telemedicine screening for prethreshold or worse ROP, and found sensitivity 100% and specificity 98%.¹⁹ Chiang et al. (163 exams from 64 infants) reported sensitivity 72%–83% and specificity 90%–99% for detection of type 2 or worse ROP, and sensitivity 85%–90% and specificity 95%–97% for detection of treatment-requiring ROP.¹³ The multicenter prospective Photo-ROP Study (300 exams from 51 infants) found sensitivity 92% and specificity 37% for detection of “clinically significant ROP”^† during weekly examinations throughout an infant’s hospital course.²⁰

Table 62.2 Diagnostic accuracy of telescreening for detection of clinically significant ROP using images captured by wide-angle camera.* Reference standard was standard indirect ophthalmoscopy

Study	Outcome measures	Sensitivity/specificity
Ells et al., 2003¹⁸	Any ROP zone I, presence of plus disease, or presence of any stage 3 ROP at any time during hospital course	1.00/0.96
Chiang et al., 2006¹³	Type 2 or worse ROP	0.72–0.83/0.90–0.99
Chiang et al., 2006¹³	Treatment-requiring ROP	0.85–0.90/0.95–0.97
Wu et al., 2006¹⁹	Prethreshold or worse ROP	1.00/0.98
Chiang et al., 2007⁸	Type 2 or worse ROP at 31–33 wks PMA	0.71–0.86/0.93–0.97
	Type 2 or worse ROP at 35–37 wks PMA	1.00/0.85–0.94
	Treatment requiring ROP at 31–33 wks PMA	NA/0.94–1.00
	Treatment requiring ROP at 35–37 wks PMA	1.00/0.81–0.94
Photo-ROP Cooperative Group, 2008²⁰	“Clinically significant ROP”	0.92/0.37
Dhaliwal et al., 2009¹⁷	Stage 3 ROP at 34 wks PMA or 4–6 wks CA	0.57/0.98
Dhaliwal et al., 2009¹⁷	Presence of plus disease	0.80/0.98
Lorenz et al., 2009²¹	Suspected treatment-requiring ROP: threshold ROP in zone II, prethreshold in zone I, or disease possibly requiring treatment but not reliably classified from images	1.00/NA
Silva et al., 2011⁹	Referral-warranted ROP: type 2 or worse	1.00/0.99
Dai et al., 2011²²	Treatment-requiring ROP	1.00/0.98

PMA, postmenstrual age; CA, chronological age; NA, not applicable.

^* RetCam; Clarity Medical Systems, Pleasanton, CA. For definition of “clinically significant ROP”, see text.

In a prospective study, Chiang et al. (248 exams from 67 infants) examined the effect of PMA and disease severity on telemedicine accuracy. At 31–33 weeks PMA, the sensitivity and specificity by three expert graders was 71%–86% and 93%–97% respectively for detection of type 2 or worse ROP, and the specificity for detection of treatment-requiring ROP was 94%–100%. At 35–37 weeks PMA, sensitivity and specificity by three graders were 91%–97% and 98%–100% respectively for detection of any ROP, 100%–100% and 85%–94% respectively for detection of type 2 or worse ROP, and 100%–100% and 81%–94% respectively for detection of treatment-requiring ROP.⁸ The finding of higher accuracy in older infants was in agreement with previous studies.¹⁵

Dhaliwal et al. (245 exams from 81 infants) conducted a masked, double-observer prospective longitudinal cohort study. Two pediatric ophthalmologists were randomized to perform examinations using either telemedicine or ophthalmoscopy. Sensitivity of telemedicine for detecting stage 3 or worse ROP was 57%, and specificity was 98% compared to ophthalmoscopy. Sensitivity for detecting plus disease was 80%, and specificity was 98%. Absolute agreement between ophthalmoscopy and telemedicine was 96% for detection of stage 3, and 97% for detection of plus disease.¹⁷ Dai et al. (422 exams from 108 infants) conducted a study in which all infants received telemedicine imaging and ophthalmoscopy by a pediatric ophthalmologist. Images were reviewed independently by a masked grader. Using ophthalmoscopy as the reference standard, sensitivity of telemedicine for detecting treatment-requiring ROP (i.e., type 1 or worse) was 100% and specificity was 98%. The positive predictive value of telemedicine for detecting treatment-requiring ROP was 86%, and the negative predictive value was 100%.²²

Scott et al. compared the accuracy of telemedicine vs ophthalmoscopy using a study design in which these two methods were performed by the same experts in 67 consecutive infants. There was substantial to near-perfect agreement between these methods, with absolute intra-grader agreement of 86% (178/206 eyes) and kappa values of 0.66–0.85 between ophthalmoscopy and telemedicine. Among the 14% (28/206 eyes) intra-expert discrepancies, some cases provided photographic evidence that ophthalmoscopy failed to recognize mild ROP that was detected by telemedicine. There were also discrepancies involving presence of zone I and plus disease, in which telemedicine may have provided theoretical advantages by allowing examiners to review their diagnoses, make more exact measurements of anatomical landmarks, and directly compare images to standard photographs.²³

Image quality

Several studies have examined the quality of images captured by nurses, ophthalmic photographers, or ophthalmologists. Ells et al. found that wide-angle images were captured successfully in 96% of examinations, and that 94% of image sets could be graded remotely. In the remaining 6%, readable photographs were obtained within 1–4 weeks.¹⁸ Wu et al. found that 79% of initial retinal images and 78% of repeated images were acceptable.¹⁹ The Photo-ROP cooperative group found that 92% of image sets were acceptable.²⁰ Chiang et al. reported that telemedicine graders reported an “unknown” diagnosis because of inadequate image quality or insufficient retinal coverage in 0%–41% of exams at 31–33 weeks PMA, and in 0%–7% of exams at 35–37 weeks PMA.⁸ Lorenz et al. reported that, among 6460 telemedicine imaging sessions conducted at five NICUs over a six-year period, nearly 98% were of adequate quality.²¹ Heavy fundus pigmentation, corneal and vitreous haze, smaller palpebral fissures, and limited dilation may be associated with decreased image quality.^8,^13,^16,¹⁹

Cost-effectiveness, speed, and satisfaction

Economic and practical factors must be considered for long-term viability of telemedicine systems. Two studies have compared the cost-effectiveness of telemedicine versus ophthalmoscopy for ROP management using decision tree models created from published results involving disease incidence, diagnostic accuracy, and treatment outcomes. One study, performed in the United States, found that telemedicine is more cost-effective than traditional ophthalmoscopy ($3193 per quality-adjusted life year [QALY] compared to $5617/QALY).²⁴ The second study, performed in the United Kingdom, modeled five possible strategies for ROP surveillance using telemedicine and ophthalmoscopy, and found that telemedicine using image capture and grading by visiting nurses (£175 per infant examined) and telemedicine using image capture by visiting nurses and image grading by remote ophthalmologists (£201/infant examined) were more cost-effective than traditional ophthalmoscopy at the NICU bedside (£321/infant examined).²⁵

Other studies have examined logistical factors such as examination time and acceptability to patients. Richter et al. found that telemedicine examinations required significantly less physician time than ophthalmoscopy (1.02–1.75 minutes per telemedicine exam versus 4.17–6.63 minutes per ophthalmoscopic exam).²⁶ Lee et al. developed and validated a survey instrument to assess attitudes toward digital imaging and telemedicine by parents, and found high acceptance of these technologies. Families did, however, report that face-to-face contact with physicians was important.²⁷

Evaluation of operational ROP telemedicine programs

Real-world telemedicine programs have been implemented in several centers, typically relying on trained neonatal technicians or nurses to capture images and transfer data for interpretation by remote ophthalmologists. Infants identified with clinically significant disease are either examined locally by an ophthalmologist, or transferred for further evaluation.

A telemedicine program involving five German NICUs has been operational since 2001. In this program, all premature infants at risk for ROP are screened with wide-angle imaging and also examined by local ophthalmologists. All suspected treatment-requiring ROP stages were detected with 100% sensitivity, and the overall positive predictive value for treatment-requiring ROP was 88.2% after 6460 examinations in 1222 infants.²¹

A similar program involving four NICUs, in which nurses are trained to capture serial images, has been used for routine telemedicine management at Stanford University since 2005.⁹ Infants felt to have referral-warranted (i.e. type 2 or worse) or treatment-requiring (i.e. type 1 or worse) ROP based on telemedicine were referred for complete ophthalmoscopic evaluation. Within one week of NICU discharge, all infants in this program received a mandatory ophthalmoscopic examination by the same retinal specialist. The sensitivity of telemedicine for identifying referral-warranted and treatment-requiring ROP was reported to be 100%, the positive predictive value of telemedicine for identifying treatment-requiring ROP was 90%, and the negative predictive value was 100%. No known cases of treatment-requiring ROP were missed, and there were no adverse outcomes such as retinal detachment.⁹

Barriers and challenges

Despite technological advances to support telemedicine for ROP management, its widespread adoption has been limited by factors such as medical licensure and lack of consistent insurance coverage and reimbursement policies.²⁸ The level of diagnostic accuracy required for implementation of real-world ROP telemedicine systems is unclear, given concerns about medicolegal liability. Furthermore, it is difficult to rigorously assess accuracy because there may be variability in the reference standard of indirect ophthalmoscopy. Capturing images with sufficient diagnostic quality may not always be practical, particularly in the peripheral retinas of younger infants, warranting re-evaluation either by repeat imaging or ophthalmoscopy. Finally, implementation of telemedicine for ROP requires approval of physicians and financial investments for new equipment and information technologies.