Background: Hospitalization for pelvic inflammatory disease (PID) treatment in young, nulliparous women often occurs despite evidence and recommendations against it. Some physicians feel that inpatient treatment lessens the risk of infertility and other PID complications in this patient group, and that available evidence showing no effect is insufficient to change practice.

Methods: We constructed a Markov decision model to estimate the cost-effectiveness of hospitalization compared to outpatient therapy for mild to moderate PID (i.e. when severe illness was absent), using it to calculate incremental cost per QALY gained for various hospitalization benefit assumptions. The baseline model examined 18-year-olds from the societal perspective over a 10 year time horizon using a 3% discount rate. Probabilities and costs were obtained from the literature, with PID outcome frequencies and utilities obtained by primary data collection. At baseline, we assumed hospitalization decreased infertility, ectopic pregnancy, and chronic pelvic pain rates equally; other scenarios, age ranges, and time horizons were examined in sensitivity analyses.

Results: If hospital treatment decreases PID complications by 10%, 20%, or 30%, the cost/QALY gained by hospitalization is $185,600, $85,100, or $53,300 compared to outpatient treatment. When hospitalization decreases complications by 10%, incremental costs/QALY are: 1) <$50,000 when hospitalization costs <$2030 (baseline $6500, baseline outpatient acute care cost $263); 2) $138,200 when the utility of infertility is 0.3 (baseline 0.76); and 3) $91,300 when 15-year-olds are considered over a 20 year time horizon. If hospital therapy decreases infertility and ectopic pregnancy rates by 10% but chronic pelvic pain rates are unaffected, then hospitalization costs >$3 million per QALY gained.

Conclusion: Hospitalization for PID treatment due to theoretical benefits in preserving future fertility in young, nulliparous women is unlikely to be economically reasonable, even if substantial (but thus far unseen) improvements in PID complication rates are assumed. Treatment effects large enough to influence the decision should have already been seen.