Calculating the economic burden of presumed microbial keratitis admissions at a tertiary referral centre in the UK

Microbial keratitis (MK) is a leading cause of global blindness, with incidence varying markedly by geography (e.g., ~11 per 100,000 in the USA to ~799 per 100,000 per year in Nepal). In the UK, MK is the most common non-surgical ophthalmic emergency (estimated 40–52 per 100,000 per year). Diagnosis often relies on clinical judgment, as corneal scrape microbiology has limited sensitivity (reported 23.7–61.5%). Management requires intensive topical antimicrobial therapy, often hourly initially, which is burdensome and can impair adherence; non-compliance with eye drops is estimated at ~36%, frequently necessitating hospitalisation for at least 24 hours and often longer. Within the NHS, revenue is driven by HRG codes and trim points; there is no specific HRG for MK admissions, leading to coding via procedures and potential income mismatches. Length of stay (LOS) is a key cost driver, while income remains fixed until trim points, incentivising early discharge. This study aimed to assess the feasibility of a data collection tool derived from the MOG clinical record form to prospectively capture epidemiology and outcomes, quantify direct inpatient costs for MK admissions, compare these to income generated through coding, and explore the effect of socioeconomic deprivation on costs.

The paper situates MK within existing literature: it highlights wide geographic variation in incidence and pathogens, high UK emergency burden, and limited sensitivity of corneal scrape microbiology (23.7–61.5%), which underpins reliance on clinical diagnosis of presumed MK. It notes substantial treatment burden from intensive drop regimens and documented non-compliance (~36%). The NHS coding structure lacks a dedicated HRG code for MK, potentially distorting reimbursement, with income determined by procedure-based coding and LOS trim points. These gaps motivate a health-economic evaluation tailored to MK admissions.

Design and setting: Retrospective cohort study of 101 consecutive MK admissions to the Birmingham and Midland Eye Centre (tertiary referral centre) from January to December 2013, using an Excel-based adaptation of the microbiology ophthalmology group (MOG) clinical record form. The tool captured four domains: demographics, clinical features, microbiology, and treatment.

Feasibility testing: The Excel data capture form was tested on ten patients by two data collectors and refined to minimise free text, enabling validated multiple selections via VBA for variables with multiple inputs (risk factors, symptoms, antibiotics, microbiology).

Disease severity stratification: Patients were graded per Keay et al. into Grades 1–4 based on lesion size/location, culture status, visual loss, and surgical intervention.

Risk factors: Ocular and systemic risk factors classified as contact lens wear; ocular surface disease (e.g., dry eye, cicatrising conjunctivitis, atopy); prior ocular surgery; trauma; diabetes mellitus; topical/systemic immunosuppression; incomplete lid closure.

Socioeconomic deprivation: Residential postcodes mapped to Index of Multiple Deprivation (IMD) 2019 deciles; dichotomised as more deprived (deciles 1–5) vs less deprived (6–10).

Direct cost of admission (COA): Calculated per patient by summing components:

• Bed-days: £273 per inpatient night (NHS reference costs), excluding staffing/treatment costs.
• Human resources: Primary data from surveys. Doctors’ survey (n=30) mean initial consultation 43.5 min (SD 14.0); hourly rate £140; unit cost per consultation £101.50. Nurses’ survey (n=4) mean time administering topical medication 4.1 min per drug per eye (SD 1.7); hourly rate £41; unit cost £2.80 per drug per eye. Assumptions: one medical consultation at admission; subsequent medical time during rounds included within bed-day cost; procedure reference costs already include staffing; every topical dose administered by a nurse, total doses multiplied by nurse unit cost.
• Interventions: Costs for corneal biopsy, corneal glue, evisceration, temporary tarsorrhaphy from NHS Improvement mean theatre runtimes and reference costs.
• Investigations: Itemised per patient for culture media (E. coli overlay, blood, chocolate, Sabouraud agar), Gram stains, contact lens testing, conjunctival swabs; costs from NHS reference costs. PCR testing outsourced at £120 per patient plus £3.30 postage.
• Drug costs: From British National Formulary. Antimicrobials costed per drop (each dose two drops), with stepped regimens: Stage 1 hourly day and night (48 drops/day); Stage 2 hourly by day, two-hourly at night (36 drops/day); Stage 3 six times daily (12 drops/day), accounting for container size and shelf life. Anti-inflammatory steroid costs computed by frequency and duration. Discharge medications: two weeks’ supply costed at discharge.

Income: Actual hospital income per admission obtained from coding department; HRG-based revenues and trim points applied by standard NHS tariffs.

Statistical analysis: Continuous variables assessed by Shapiro–Wilk (non-normal), summarised as medians (IQRs) and means for cost comparability. Mann–Whitney U and Fisher’s exact tests compared patient factors by severity grades. Wilcoxon signed-rank test compared COA vs income per patient. Association between LOS and COA–income discrepancy analysed with regression using log10-transformed LOS to fit a logarithmic trend; model residuals examined; cost-neutral LOS (discrepancy=0) estimated. Analyses performed in STATA and SPSS; significance p<0.05.

• Cohort and severity: 101 admissions; median age 59 years (IQR 38–73). Most had severe disease: Grade 4 = 80.2%. Median LOS 7 days (IQR 5–10). No corneal transplantation or amniotic membrane grafting required.
• Coding availability: Hospital coding data available for 97/101 (96.0%); 4 uncoded admissions excluded from cost–income analyses.
• Direct costs (COA): Median COA £2,855 per patient (IQR £2,018–£4,057); mean £3,681. Total COA across 97 patients £357,075. Cost composition: bed-days 68.0% of total; human resource 20.3%.
• Income: Total income £252,116 (n=97); median income £2,124 (mean £2,599).
• Deficit: Overall deficit £104,960 for the year (median deficit £754 per patient). COA significantly exceeded income (Wilcoxon signed-rank test; p-value not explicitly stated, but deficits reported).
• Length of stay (LOS): Cost deficit increased significantly with LOS (p < 0.001). Patients with short LOS generated positive margins; model-estimated cost neutrality at 4.8 days LOS.
• Socioeconomic deprivation: Greater deprivation (IMD deciles 1–5) was associated with a significantly higher cost deficit (exact statistics not reported in excerpt, direction significant).
• Human resource unit costs: Medical consultation unit cost £101.50 (43.5 min at £140/hour). Nursing administration cost £2.80 per drug per eye (4.1 min at £41/hour).
• Treatment dosing framework: Stepped antimicrobial regimens (48, 36, then 12 drops/day) informed dose-based drug and nursing time costs.

The study demonstrates that inpatient management of presumed MK at a UK tertiary centre generates substantial direct costs that exceed HRG-based income, largely driven by length of stay. Because current reimbursement is fixed up to HRG trim points, earlier discharge after the initial sterilisation phase can markedly reduce costs without decreasing income, creating potential for cost neutrality or surplus at shorter LOS. The significant association between higher socioeconomic deprivation and greater cost deficits suggests that patients from more deprived areas may require longer or more resource-intensive inpatient care, compounding financial pressures on hospitals serving these populations. The feasibility testing indicates that a MOG-derived electronic tool can standardise and streamline collection of clinical, microbiological, treatment, and cost data, enabling more accurate costing and prospective epidemiological surveillance. In the context of the current lack of an MK-specific HRG code, such granular data could support service planning, inform coding practices, and advocate for more appropriate tariffs reflecting MK’s resource intensity. Overall, findings address the study aims by quantifying direct inpatient costs, identifying LOS as the principal modifiable cost driver, and highlighting socioeconomic factors influencing cost burdens.

An Excel-based, MOG-derived data collection tool feasibly captures the key clinical and economic data needed to quantify direct inpatient costs for MK admissions. In this cohort, direct costs substantially exceeded income, with LOS being the dominant driver of the deficit; cost neutrality was achieved at an LOS of approximately 4.8 days. Clinical protocols should support timely discharge once the infection is sterilised and patients can self-administer drops safely. Future work should develop a multiuser, multisite platform for robust prospective data capture and a national registry, and expand to include indirect costs (e.g., treatment burden, visual morbidity, quality of life) to fully characterise disease burden and inform appropriate reimbursement structures.

• Retrospective, single-centre design (2013 admissions) may limit generalisability and temporal relevance.
• Missing coding data for 4/101 admissions prevented inclusion in cost–income analyses.
• Human resource time estimates derived from small surveys (n=30 doctors; n=4 nurses) at a single institution may introduce measurement error and limit external validity.
• Costing relied on national reference costs and assumptions (e.g., all topical doses nurse-administered in hospital), which may not reflect practice variations.
• Lack of an MK-specific HRG code may lead to income estimates that do not fully reflect delivered care complexity.
• Indirect costs (patient/caregiver burden, productivity loss, quality of life) were not captured.