Clear communication between clinicians and patients isn't a soft skill — it's a measurable determinant of whether people actually follow through on their care. When the letters summarizing a clinical encounter are written at college level, a substantial portion of patients may leave confused, disengaged, or less likely to adhere to treatment. A decade-long dataset from one of the world's busiest eye hospitals now puts hard numbers on a long-suspected problem.

Analyzing over 4.6 million outpatient letters generated for nearly 805,000 patients at Moorfields Eye Hospital between 2013 and 2025, researchers applied multiple validated readability instruments — the Flesch Reading Ease score, the SMOG Index, and the Automated Readability Index — alongside linguistic complexity measures such as lexical density and type-token ratio. The median Flesch Reading Ease score landed at 51.1, a range conventionally associated with college-level text. Recommended readability for patient-facing health materials typically targets a score of 60 or above, corresponding roughly to an eighth-grade reading level. Fully 95.3% of letters fell below that threshold, and the pattern held consistently across all 17 ophthalmology subspecialties analyzed, with statistically significant variation between services but no strong link to patient sociodemographic characteristics.

This work matters beyond ophthalmology. Readability failures in clinical correspondence are a systemic healthcare communication issue, and the scale of this dataset lends unusual statistical credibility to what smaller studies have suggested for years. The implications are notable given that roughly one in six adults in the UK reads at or below the literacy level of an average 11-year-old. This study is observational and specialty-specific, so causal links to downstream outcomes like missed appointments or medication errors cannot be drawn directly. It also does not assess comprehension or patient-reported experience. Nevertheless, the longitudinal span — 12 years — and sheer volume make this one of the most robust readability audits in clinical literature, and it adds measurable evidence to calls for AI-assisted plain-language tools in electronic health record systems.