A store-and-forward telemedicine for retinopathy of prematurity screen: is it cost-effective in Thailand?

A cost–utility analysis was conducted to compare the costs and outcomes of digital photographic ROP screening using the RetCam, an example of store-and-forward telemedicine, with interpretation by geographically remote experts with the current practice whereby general ophthalmologists determine whether a referral is needed. This study was approved by the Khon Kaen University Ethics Committee (approval Nos. HE551325 and HE57118).

The study was conducted in light of the Thai context using the viewpoint of the providers and society, with hypothetical birth cohorts. A decision tree model to compare the costs and outcomes in terms of blindness prevention and quality-adjusted life years (QALYs) of alternative ROP screenings and diagnosis was developed in Microsoft Excel software (Microsoft Office 2007). Figure 1 illustrates the screening pathway using telemedicine compared with the current practice.

Figure 1

For the telemedicine method, trained technicians use RetCam to capture photographs, which are then sent digitally to expert ophthalmologists for interpretation. In these files, the capture of the fundus may or may not be interpretable depending on the child. In the case where the photo is interpretable, a positive case will be referred to experts for further care and treatment, while a negative case will not be referred. For noninterpretable cases, all newborns are referred for further investigation. The current practice—the conventional ROP screening method—requires general ophthalmologists to conduct eye examinations using an indirect ophthalmoscope to inspect the fundus of preterm infants. Referred infants are further investigated for ROP by experts in a higher-level hospital to confirm the disease and offer treatment accordingly.

A lifetime time horizon was used to estimate the long-term costs and outcomes of infants becoming blind. All future costs and outcomes were discounted at a rate of 3% per annum. For intercountry comparisons, costs were converted into international dollars ($IS) using the purchase power parity exchange rate of $IS1 = 17.659 THB (Thai baht) [9]. All costs were adjusted to 2013 values using the general consumer price index [10]. A probabilistic sensitivity analysis was conducted using a second-order Monte Carlo simulation with 1,000 repetitions, which incorporates statistical uncertainty into the model.

Key parameters used in the decision models are summarized in Table 1.

A prospective clinical trial was conducted from September to December 2013 to assess the diagnostic accuracy of RetCam in a northeast region of Thailand (Thai Clinical Trials Registry Identification No. TCTR20130911001). The mother, father, or legally authorized representative of the infant provided written informed consent for the pediatric patient participation in the study. A total of 100 infants were enrolled over the duration of 3 months. Nine of 100 infants were diagnosed with ROP requiring treatment (ROPRT) according to the criteria set by the ETROP study [11]. Of these 9 infants, 8 had bilateral disease.

The sensitivity of the RetCam in detecting ROP-RT was 1.00 (95% CI 0.70, 1.00) and specificity was 0.98 (95% CI 0.92, 0.99). The fundus images of 2 infants were not able to be interpreted because of their poor quality so they were referred to an expert ophthalmologist; this resulted in neither of the infants requiring any treatment. From interpreting the fundus images of the remaining 98 infants, 11 were categorized with ROP-RT and 87 of the 98 did not require any treatment.

The other screening method was indirect ophthalmoscopy conducted by local ophthalmologist, which showed a sensitivity of 0.89 (95% CI 0.57, 0.98) and specificity of 0.93 (95% CI 0.86, 0.97) for the detection of ROP-RT. Based on this method, 14 of the 100 infants were referred to meet expert ophthalmologists. Of the 86 remaining infants that did not require any treatment, 1 of was diagnosed at high risk for prethreshold ROP via the criterion standard for ROP diagnosis, and an assessment by an expert ophthalmologist using indirect ophthalmoscopy. This infant required laser photocoagulation treatment.

The costs of the telemedicine program were collected from a prospective clinical trial in order to determine the diagnostic accuracy of the RetCam. Direct medical costs were identified from a database, chart reviews and expert opinions, while direct non-medical costs were collected from the clinical trial using a cost questionnaire and cost diary.

For the telemedicine program, the annual capital costs for machinery were calculated by dividing the total costs by the annuity factor, when scrap value equals to 0. The annuity factor can be calculated using the following formula:

Annuityfactor=1r−1r1+nn,$$Annuity\,factor = \left( {\frac{1}{r}} \right) - \left( {\frac{1}{{r\left( {1 + n} \right)n}}} \right),$$

where r represents the discount rate and n denotes the number of years. An annual discount rate of 3% was used. The costing guideline recommended that medical machinery should be estimated at around 8 years, with variation between 5 and 10 years. The direct medical costs of telemedicine and the current practice are shown in Table 2.

Based on the natural course of ROP, when an infant does not receive any treatment within an appropriate timeframe, it has a very high probability of becoming blind. Even when an appropriate treatment such as the use of a laser is available, the probability of becoming blind is still 9.82% [4]. Blindness can cause mortality and morbidity, resulting in a lower quality of life.

In determining utility, values for monocular and binocular blindness were referenced from published literature [12]. For those with favorable outcomes, the utility is equal to 1. Subsequently, QALYs were calculated by adjusting the hazard ratio of visual impairment on mortality. The hazard ratio of monocular blindness compared with no blindness was 1.23 (95% CI 1.16, 1.31) and the hazard ratio of bilateral blindness was 1.54 (95% CI 1.28, 1.86) [13]. Data from the WHO Global Health Observatory Data Repository were used to calculate life years [14].

There were several assumptions made regarding clinical practice and costs. For telemedicine, it is assumed that nurses prepare infants in a group of approximately 10 cases per week in one day. For the current method, the preparation ratio of nurses and infants is one-to-one and examinations can be conducted on any working day when a pediatrician sends a request to the ophthalmologist for ROP screening.

Labor costs for an expert ophthalmologist include the time they take with individual infants, including the time to wait for pupil dilatation and examination using a conventional method using indirect ophthalmoscopy. However, for ROP screening via the RetCam, the expert ophthalmologist only needs time to interpret the fundus images. Because there are no data regarding the repair and maintenance costs of the machine, we assumed that these costs account for 10 percent of the total cost.