Research Letter
Initial Antibiotic Selection Strategy and Subsequent Antibiotic Use—Insights From the INSPIRE Trials
Shruti K. Gohil, Edward Septimus, Ken Kleinman, et al
JAMA 2025;334;(12):1107-1109
doi:10.1001/jama.2025.11256
Four cluster-randomized trials have demonstrated that a stewardship bundle including real-time computerized provider order entry (CPOE) prompts safely reduced initial extended-spectrum antibiotic use by 17% to 35% among adult patients hospitalized with common infections who were at low risk of infection with a multidrug-resistant organism (MDRO).1-4 The CPOE prompts triggered during the first 3 days of hospitalization, before pathogens are typically identified, and provided clinicians with patient-specific MDRO risk estimates. We evaluated whether initial reductions in extended-spectrum antibiotics were sustained during the remaining hospitalization.
The INSPIRE (INtelligent Stewardship Prompts to Improve Real-Time Empiric Antibiotic Selection) trials (NCT05423756, NCT05423743, NCT03697070, and NCT03697096) were 4 cluster-randomized trials evaluating antibiotic selection in non–critically ill adults (≥18 years) hospitalized with pneumonia, urinary tract infection (UTI), abdominal infection, and skin or soft tissue (SST) infection.1-4 The pneumonia and UTI trials (implemented simultaneously) involved 59 hospitals, an 18-month baseline period (April 2017 through September 2018), and a 15-month intervention period (April 2019 through June 2020); the abdominal and SST infection trials (implemented simultaneously) involved 92 hospitals and 12-month baseline (January 2019-December 2019) and intervention (January 2023-December 2023) periods. Each trial compared rates of empiric extended-spectrum antibiotic days of therapy in hospitals where physicians received CPOE prompts recommending standard-spectrum instead of extended-spectrum antibiotics during the first 3 hospital days (empiric period) for patients with less than 10% risk of MDRO infection vs hospitals where physicians continued usual practice (routine stewardship). Intervention hospitals additionally received education and prescribing feedback.
We evaluated extended-spectrum antibiotic days of therapy after the third hospital day (the postempiric period) and subsets of vancomycin and antipseudomonal therapy among patients hospitalized for 4 or more days, calculated as the sum of different extended-spectrum antibiotics received per patient daily. This study was approved by the Harvard Pilgrim Health Care Institute institutional review board; individual informed consent was waived.
Unadjusted as-randomized outcomes were analyzed using disease-specific generalized linear mixed-effects models assessing differences in postempiric extended-spectrum days of therapy between intervention and baseline periods across study groups, clustering admission data within patient, hospital, and period within hospital. Analyses adjusted for demographics and comorbidities were performed. Separate analyses were conducted for each INSPIRE trial. Analyses used SAS version 9.4 (SAS Institute) or R version 4.2.3 (R Foundation for Statistical Computing).
Including all 4 INSPIRE trials, 413 901 adults were hospitalized more than 3 days. Mean (SD) age was 66.0 (17.8) years, and 44.1% were male. During the baseline period across the 4 trials, 38% to 44% of all antibiotic doses were given during the empiric period, and of patients receiving extended-spectrum antibiotics, 79% to 94% were initiated in the empiric period. See previous articles for further details.1-4
Compared with routine stewardship, postempiric (beyond day 3) extended-spectrum days of therapy in the CPOE hospitals decreased by 22% (95% CI, 16%-28%; P < .001), 11% (95% CI, 8%-15%; P < .001), 23% (95% CI, 17%-27%; P < .001), and 23% (95% CI, 15%-30%; P < .001) in the pneumonia, UTI, abdominal, and SST infection trials, respectively (Table). Similar reductions were observed for antipseudomonal subsets. Point estimate reductions in vancomycin were observed in all 4 trials, but this difference reached statistical significance only for pneumonia and skin and soft tissue infection. Adjusted analyses were similar. Of reductions achieved in empiric extended-spectrum antibiotic use, 65% to 84% were maintained through the remainder of the hospitalization (Figure); results are based on unadjusted relative rate ratios from difference-in-differences generalized linear mixed-effects models accounting for clustering within hospitals and period within hospital.
Table. INSPIRE Trials Cohorts and Primary and Secondary Outcomes During the Postempiric Period, As-Randomized Analysis
| Computerized provider order entry bundle | Routine stewardship | Total No. of patients | P value | |||||
|---|---|---|---|---|---|---|---|---|
| Baseline | Intervention | No. of hospitals | Baseline | Intervention | No. of hospitals | |||
| INSPIRE trial cohortsa | ||||||||
| Pneumonia | 23 291 | 19 424 | 30 | 21 938 | 18 631 | 29 | 83 284 | |
| Urinary tract infection | 31 156 | 24 218 | 30 | 32 359 | 23 777 | 29 | 111 510 | |
| Abdominal infection | 30 851 | 33 356 | 46 | 30 574 | 36 042 | 46 | 130 823 | |
| Skin and soft tissue infection | 20 763 | 22 192 | 46 | 21 782 | 23 547 | 46 | 88 284 | |
| Days-of-therapy outcomes during postempiric period | Baseline, days-of-therapy raw rateb | Intervention, days-of-therapy raw rateb | Rate ratio (95% CI)c | Baseline, days-of-therapy raw rateb | Intervention, days-of-therapy raw rateb | Rate ratio (95% CI)c | Overall rate ratio difference-in-differencesd | |
| Pneumonia | ||||||||
| Extended-spectrum days of therapy | 426.0 (48 529/124 266) | 252.9 (28 810/107 532) | 0.74 (0.70-0.78) | 409.9 (46 703/113 928) | 340.7 (38 811/98 416) | 0.95 (0.90-1.00) | 0.78 (0.72-0.84) | <.001 |
| Vancomycin days of therapy | 163.7 (18 649/124 266) | 94.7 (10 787/107 532) | 0.67 (0.63-0.70) | 157.4 (17 935/113 928) | 121.7 (13 862/98 416) | 0.79 (0.75-0.83) | 0.85 (0.79-0.91) | <.001 |
| Antipseudomonal days of therapy | 235.9 (26 870/124 266) | 140.6 (16 022/107 532) | 0.75 (0.72-0.77) | 228.0 (25 970/113 928) | 197.0 (22 444/98 416) | 0.99 (0.96-1.03) | 0.75 (0.72-0.79) | <.001 |
| Urinary tract infections | ||||||||
| Extended-spectrum days of therapy | 280.9 (44 101/150 121) | 182.5 (28 662/121 851) | 0.96 (0.93-0.98) | 312.7 (49 100/157 010) | 235.8 (37 016/115 151) | 1.08 (1.05-1.11) | 0.89 (0.85-0.92) | <.001 |
| Vancomycin days of therapy | 87.6 (13 761/150 121) | 55.5 (8710/121 851) | 0.84 (0.79-0.89) | 95.5 (14 996/157 010) | 67.0 (10 527/115 151) | 0.91 (0.86-0.97) | 0.92 (0.84-1.01) | .07 |
| Antipseudomonal days of therapy | 149.0 (23 388/150 121) | 95.1 (14 924/121 851) | 0.91 (0.87-0.95) | 173.3 (27 209/157 010) | 132.2 (20 761/115 151) | 1.08 (1.04-1.12) | 0.85 (0.80-0.90) | <.001 |
| Abdominal infections | ||||||||
| Extended-spectrum days of therapy | 353.4 (52 633/149 466) | 262.4 (39 080/164 817) | 0.73 (0.70-0.76) | 351.3 (52 316/148 913) | 388.6 (57 872/185 677) | 0.94 (0.90-0.99) | 0.77 (0.73-0.83) | <.001 |
| Vancomycin days of therapy | 75.8 (11 286/149 466) | 58.0 (8635/164 817) | 0.74 (0.67-0.81) | 73.7 (10 971/148 913) | 76.2 (11 353/185 677) | 0.83 (0.76-0.90) | 0.89 (0.78-1.01) | .06 |
| Antipseudomonal days of therapy | 241.9 (36 028/149 466) | 171.8 (25 580/164 817) | 0.66 (0.64-0.68) | 240.3 (35 782/148 913) | 271.4 (40 412/185 677) | 0.95 (0.92-0.98) | 0.70 (0.67-0.73) | <.001 |
| Skin and soft tissue infections | ||||||||
| Extended-spectrum days of therapy | 265.8 (33 143/113 252) | 209.6 (26 132/130 105) | 0.76 (0.71-0.81) | 285.2 (35 555/124 669) | 295.9 (36 885/145 709) | 0.99 (0.92-1.06) | 0.77 (0.70-0.85) | <.001 |
| Vancomycin days of therapye | 308.1 (38 411/113 252) | 290.9 (36 270/130 105) | 0.80 (0.78-0.82) | 315.9 (39 379/124 669) | 329.1 (41 028/145 709) | 0.89 (0.86-0.91) | 0.90 (0.86-0.93) | <.001e |
| Antipseudomonal days of therapy | 251.0 (31 295/113 252) | 196.3 (24 472/130 105) | 0.70 (0.65-0.76) | 268.9 (33 527/124 669) | 276.6 (34 478/145 709) | 0.96 (0.88-1.03) | 0.73 (0.66-0.82) | <.001 |
Figure. Proportion of Reductions in Empiric Extended-Spectrum Days of Therapy Achieved in the INSPIRE Trials1-4
aEmpiric period: hospital days 1 to 3; postempiric: hospital day 4 to discharge.
bCalculated as percentage reduction in extended-spectrum days of therapy during the postempiric period divided by reductions in the empiric period.
Patient-specific CPOE prompts to reduce empiric extended-spectrum antibiotic use resulted in sustained reductions for the entire hospital stay. Up to 84% of reductions in extended-spectrum antibiotic use achieved during the empiric period persisted in the postempiric period, confirming clinician tendency to maintain initial therapy choices. Limitations include inability to assess patient- or physician-related factors associated with postempiric period reductions. These findings suggest that investing in stewardship for initial antibiotic selection, rather than primarily focusing on de-escalating antibiotics once started, would reduce unnecessary extended-spectrum antibiotics for millions of patients in US hospitals annually.5,6
