Early phosphate supplementation after newly detected hypophosphatemia in older ICU patients: a target trial emulation

Hypophosphatemia is common in critical illness, but whether initiating replacement promptly after a newly available low result improves clinical outcomes is uncertain. We used target trial emulation to estimate a per-protocol contrast in older intensive care unit (ICU) patients. We used MIMIC-IV version 3.1 to emulate a pragmatic target trial among patients aged 65 years or older with a first ICU stay and a first qualifying blood, serum, or plasma phosphate concentration < 2.5 mg/dL obtained after ICU admission and within 48 h. Time zero was result availability. We compared initiation of eligible systemic phosphate within 24 h with no initiation; route, formulation, and dose were not standardized, treatment after 24 h was unrestricted, and doses were summarized in the units recorded in the source data. Post-dose phosphate was described among monitored treated patients. Per-protocol contrasts were estimated using clone-censor inverse-probability weighting. Among 7,535 patients, 2,764 (36.7%) received phosphate within 24 h. Estimated 28-day mortality was 14.8% (95% CI 13.5 to 16.2) under phosphate and 16.8% (15.7 to 18.0) under no phosphate (risk difference − 2.0% points, − 3.6 to − 0.3; risk ratio 0.88, 0.79 to 0.98; weighted Cox HR 0.87, 0.78 to 0.98). ICU-free days did not differ clearly, whereas hospital-free days were modestly lower under phosphate (mean difference − 0.57 days, − 1.01 to − 0.13). Hyperphosphatemia and hypocalcemia increased modestly; new renal replacement therapy was rare. The 0–6 h versus 6–24 h comparison was exploratory. Among treated patients with a repeat phosphate measurement within 24 h (90.9%), 56.7% reached ≥ 2.5 mg/dL. Initiation of heterogeneous systemic phosphate supplementation within 24 h was associated with a modest estimated reduction in 28-day mortality under the assumptions of the emulation, but ICU-free days did not clearly improve and hospital-free days were modestly lower. Small biochemical safety trade-offs and plausible residual confounding limit clinical interpretation. Timing findings were exploratory and do not establish an optimal window. Randomized trials should test standardized, monitored strategies.