A higher failure rate is seen with tube shunt implantation than with trabeculectomy, according to data extracted from electronic health records using an automated pipeline. Researchers wanted to assess glaucoma surgery outcomes over the long term by using an automated process to obtain results from electronic health data. The research was conducted in retrospect at a single research institution. EHR procedure codes were used to identify patients with a trabeculectomy, Ex-PRESS shunt, Baerveldt tube shunt, or an Ahmed tube shunt inserted between 2009 and 2018. Using a natural language processing pipeline that has been previously verified, investigators were able to extract patient characteristics from both structured and unstructured fields.
There were a total of 711 glaucoma procedures were performed on 512 individuals, including 287 trabeculectomies, 47 Ex-PRESS shunts, 274 Baerveldt tubes, and 103 Ahmed tubes. The Median follow-up time was 359 days. Overall, the mean initial intraocular pressure was 24.4 mm Hg (SD 10.9), and 72.9% of patients were taking more than or equal to 3 drugs. Surgery with a tube shunt carried a higher risk of failure than trabeculectomy (Baerveldt: Hazard Ratio (HR) 1.44, 95% CI 1.02 to 2.02; Ahmed: HR 2.01, 95% CI 1.28 to 3.17). Previous glaucoma surgery was associated with increased failure (≥2 previous surgeries: HR 2.74, 95% CI 1.62 to 4.64), similar to the decreased number of pre-medications (<3 medications: HR 2.96, 95% CI 2.12 to 4.13) and male sex (HR 1.40, 95% CI 1.03 to 1.90). At 1 year, the IOP of those who had undergone a tube shunt was 2.53 mm Hg (P=0.002) greater than that of those who had undergone a trabeculectomy.
In addition, more failures occurred after Baerveldt and Ahmed tube shunt implantation than after trabeculectomy. Factors that increased the likelihood of failure were a lower number of drugs at baseline, a history of glaucoma surgery, and being male. These findings highlight the value of implementing an informatics pipeline on electronic health records to probe important clinical topics with data from the real world.
