Introduction: Healthcare-associated infections (HAIs) caused by carbapenem-resistant gram-negative bacteria are common in Indian tertiary hospitals. Few Indian studies have tested combined antimicrobial stewardship (AMS) and infection prevention and control (IPC) programmes against infection, consumption and resistance outcomes together. Objective: To assess whether a structured, microbiologist-led AMS and IPC intervention reduces HAI incidence, antimicrobial consumption and antimicrobial resistance. Design, setting, and participants: Pre-post quasi-experimental study with interrupted time-series (ITS) analysis in 3 adult intensive care units (ICUs) and 14 adult wards of a 650-bed tertiary care teaching hospital in India. All adults admitted for 48 hours or longer during the preintervention period (January–December 2024; 18,950 admissions; 130,272 patient-days) and the postintervention period (April 2025–March 2026; 19,438 admissions; 130,852 patient-days) were included. January–March 2025 was the implementation phase and was excluded. Intervention: A bundled programme with 4 components: prospective audit and feedback with a 48–72-hour antibiotic time-out; restriction of carbapenems, polymyxins and other Reserve-group agents; an antibiogram-based treatment and surgical prophylaxis policy; and a multimodal IPC package covering hand hygiene, device bundles and contact precautions. Main outcomes and measures: The primary outcome was overall HAI incidence per 1000 patient-days. Secondary outcomes were device-associated infection rates, surgical site infection (SSI), antimicrobial days of therapy (DOT) per 1000 patient-days, resistance proportions and multidrug-resistant organism (MDRO) incidence, and in-hospital mortality. Results: Hand hygiene compliance rose from 45.8% to 78.4%. Overall, HAI incidence fell from 5.97 to 3.48 per 1000 patient-days (incidence rate ratio [IRR], 0.58; 95% CI, 0.52–0.66; P < .001). The ITS level change was IRR 0.63 (95% CI, 0.48–0.82), and the case-mix–adjusted IRR was 0.64 (95% CI, 0.49–0.84). Central line–associated bloodstream infection fell by 48% (IRR, 0.52; 95% CI, 0.36–0.74), catheter-associated urinary tract infection by 41% and ventilator-associated pneumonia by 37%. SSI fell from 4.6% to 2.7% (risk ratio [RR], 0.59; 95% CI, 0.49–0.71). Total DOT fell by 17.2% and carbapenem DOT by 35.6%. Carbapenem resistance declined in Klebsiella pneumoniae (67.0% to 59.7%; RR, 0.89; 95% CI, 0.80–0.99), Acinetobacter baumannii (88.7% to 80.7%) and Pseudomonas aeruginosa (42.2% to 29.3%). Incidence of hospital-onset carbapenem-resistant Enterobacterales fell by 31% (IRR, 0.69; 95% CI, 0.55–0.88). Resistance in community-onset outpatient Escherichia coli, a nonequivalent control outcome, was unchanged (58.7% vs 58.8%). Mortality did not differ (6.7% vs 6.4%; RR, 0.96; 95% CI, 0.89–1.03). Conclusions and relevance: A combined AMS and IPC intervention was associated with fewer HAIs, lower carbapenem use and lower carbapenem resistance within 12 months, with no rise in mortality. Multicenter controlled ITS studies should test whether these effects hold and persist.
Bacterial antimicrobial resistance (AMR) was directly responsible for an estimated 1.27 million deaths and associated with 4.95 million deaths worldwide in 2019.¹ WHO surveillance data for 2023 show that about 1 in 6 laboratory-confirmed bacterial infections worldwide were resistant to antibiotics. In the WHO South-East Asia Region, the figure was about 1 in 3.²
Healthcare-associated infections (HAIs) are a main route by which resistant organisms reach patients. In low- and middle-income countries the pooled HAI prevalence is about twice that of high-income settings. ³ Indian surveillance confirms a high burden. In the International Nosocomial Infection Control Consortium (INICC) network, adult and paediatrics ICUs reported 5.1 central line–associated bloodstream infections (CLABSIs) and 9.4 ventilator-associated pneumonias (VAPs) per 1000 device-days.⁴ In the national HAI surveillance network of 26 tertiary hospitals, 72% of Klebsiella bloodstream infections and 77% of Acinetobacter bloodstream infections were carbapenem resistant.⁵ The ICMR AMR Research and Surveillance Network continues to report rising carbapenem resistance in K pneumoniae and A baumannii, with the highest levels in ICUs.⁶ High antibiotic consumption, unregulated access and weak infection control drive these trends.⁷
Two strategies address this problem from different directions. Antimicrobial stewardship (AMS) reduces selection pressure by improving antibiotic choice, dose and duration.⁸˒⁹ Infection prevention and control (IPC) reduces transmission and the number of infections that need treatment at all.¹⁰ A Cochrane review of 221 studies found that hospital stewardship interventions increased guideline-compliant prescribing and shortened treatment without increasing mortality.¹¹ A meta-analysis of 32 studies found that stewardship reduced colonization and infection with multidrug-resistant gram-negative bacteria by 51%. The effect was largest when stewardship was combined with IPC measures, particularly hand hygiene.¹² Landmark single-component IPC programmes have also cut HAIs substantially.¹³˒¹⁴
The evidence from India is thinner. A meta-analysis of Asian stewardship programmes found consistent reductions in antibiotic use but no pooled effect on HAI, except in programmes that included infection control or hand hygiene.¹⁵ Indian studies have mostly reported process and cost outcomes from single stewardship programmes.¹⁶ National policy documents, ICMR stewardship guidelines and the National Action Plan on AMR 2.0 (2025–2029) all call for hospital-level AMS and IPC programmes.¹⁷⁻¹⁹ Few evaluations, however, have measured HAI incidence, antimicrobial consumption and resistance together. Fewer still have used designs that separate an intervention effect from secular trends. ²⁰
We therefore evaluated a structured, microbiologist-led AMS and IPC intervention in a tertiary teaching hospital. We used a pre-post design with interrupted time-series analysis and a nonequivalent dependent variable.
Objectives. The primary objective was to measure the change in overall HAI incidence per 1000 patient-days. The secondary objectives were to measure changes in device-associated infections, surgical site infection (SSI), antimicrobial consumption, resistance proportions in key pathogens, MDRO incidence and in-hospital mortality.
Hypothesis. We hypothesised that the intervention would reduce overall HAI incidence by at least 25%, with no increase in mortality.
Study design and reporting
This was a single-centre, pre-post quasi-experimental study with interrupted time-series (ITS) analysis of monthly data. It included a nonequivalent dependent variable to detect secular confounding.²⁰ The report follows the TREND statement, the ORION statement for intervention studies of nosocomial infection, and SQUIRE 2.0.²¹⁻²³
Setting
The study took place in a 650-bed public tertiary care teaching hospital in India. The hospital has about 26,000 adult admissions a year and an NABL-accredited microbiology laboratory. The study units were 3 adult ICUs with 60 beds (medical 24, surgical 20, emergency and trauma 16) and 14 adult medical and surgical wards with 360 beds. Paediatric, neonatal and obstetric units were excluded because they have a different case mix and a separate neonatal IPC programme.
Before the intervention, the hospital had a Hospital Infection Control Committee and 2 infection control nurses. Prospective HAI surveillance had run since July 2023, using the same definitions and the same team as during the study. There was no stewardship team, no restricted-antimicrobial policy and no antibiogram-based treatment guideline.
Study periods
Participants
All patients aged 18 years or older admitted to a study unit for 48 hours or longer were eligible. Day-care admissions and stays shorter than 48 hours were excluded. Infections present on admission (onset before hospital day 3) were not counted as HAIs. Each patient contributed patient-days for every day spent in a study unit. Device-days were counted daily at 08:00 by ICU nurses.
Intervention
A multidisciplinary AMS team ran the programme. It was led by a clinical microbiologist and included an intensivist, an internal medicine physician, a clinical pharmacologist, a pharmacist and the senior infection control nurse. The design followed the ICMR Antimicrobial Stewardship Program Guideline and the WHO core components for IPC programmes.¹⁰˒¹⁸ Table 1 lists the components.
Outcome definitions
HAIs were identified by prospective surveillance by trained infection control nurses. The microbiologist validated each case. Definitions followed the CDC National Healthcare Safety Network (NHSN) Patient Safety Component Manual, January 2024, as adapted for the Indian HAI surveillance network. ⁵˒³⁰
Microbiology
Isolates were identified by MALDI-TOF mass spectrometry or VITEK 2 (bioMérieux). Antimicrobial susceptibility testing used VITEK 2 AST cards with disc diffusion confirmation, interpreted against the CLSI M100 34th edition breakpoints for the whole study.³³ Colistin susceptibility was tested by broth microdilution. Quality control used ATCC reference strains, and the laboratory took part in external quality assessment throughout. Breakpoints and testing platforms did not change during the study.
Sample size
We assumed a baseline HAI incidence of 6.0 per 1000 patient-days and set out to detect a 25% reduction (to 4.5 per 1000 patient-days) with 80% power and two-sided α = .05. This needed about 36,600 patient-days per period. Each period was expected to provide about 130,000 patient-days. For ITS analysis we used 12 monthly data points before and after the intervention, the minimum recommended for segmented regression.³⁴
Statistical analysis
Baseline characteristics were compared using standardised mean differences (SMDs). An SMD above 0.10 was taken to indicate meaningful imbalance. Rates were compared as incidence rate ratios (IRRs) with Wald 95% CIs on the log scale and exact conditional binomial P values. Proportions were compared as risk ratios (RRs) using the χ² test, or Fisher exact test when any expected cell count was below 5.
The primary ITS analysis used segmented quasi-Poisson regression of monthly HAI counts, with the logarithm of patient-days as an offset.³⁵˒³⁶
The adjusted model added monthly mean Charlson Comorbidity Index and ICU share of patient-days. Device utilisation was not adjusted for, because reducing device use was part of the intervention and so lies on the causal pathway. Antimicrobial consumption was modelled with segmented linear regression and Newey-West standard errors (lag 3).
We ran 2 sensitivity analyses: negative binomial regression in place of quasi-Poisson, and inclusion of the implementation months in the postintervention period. Monthly denominators were complete. Isolates without susceptibility results were excluded (less than 1% of isolates). Secondary outcomes were not adjusted for multiple comparisons and are interpreted as exploratory.
All tests were two-sided with α = .05. Analyses were run in Python 3.13 with statsmodels 0.15 and SciPy
Table 1. Components of the antimicrobial stewardship and infection prevention intervention
|
Component |
Content |
Frequency |
Lead |
|
Governance |
AMS committee reporting to the Medical Superintendent; joint AMS–HICC review of unit dashboards |
Monthly |
Microbiologist (chair) |
|
Treatment and prophylaxis policy |
Empirical treatment guideline built from the 2024 cumulative antibiogram; single-dose surgical prophylaxis given within 60 min of incision and stopped by 24 h |
Policy issued Jan 2025; antibiogram updated yearly |
AMS team |
|
Restriction |
Carbapenems, polymyxins, tigecycline, ceftazidime-avibactam and linezolid released for 24 h on a justification form; continuation needs AMS approval |
Daily |
Microbiologist, pharmacist |
|
Prospective audit and feedback (PAF) |
Review of all restricted agents and of every ICU antibiotic course past 72 h; written recommendation in the case record and a phone call to the treating team |
Weekdays |
AMS team |
|
Antibiotic time-out |
Printed 48–72 h checklist in the drug chart: diagnosis confirmed, cultures reviewed, de-escalate, switch to oral, stop date set |
Every course |
Treating physician |
|
Diagnostic stewardship |
Paired blood cultures before the first dose; no urine culture from catheterised patients without symptoms; same-day phoned reports of positive blood cultures and MDRO flags |
Continuous |
Microbiology laboratory |
|
Hand hygiene |
WHO multimodal strategy: alcohol-based handrub at every bedside, training, direct observation of the 5 moments, unit feedback and reminders²⁴ |
Observation audits monthly |
IPC nurses |
|
Device and SSI bundles |
CLABSI, CAUTI, VAP and SSI bundles adapted from current SHEA/IDSA/APIC compendium recommendations, with daily ICU checklists and a daily review of whether each device is still needed²⁵⁻²⁸ |
Daily checklists; monthly audits |
ICU nurse champions |
|
MDRO control |
Contact precautions and cohorting for CRE and CRAB; rectal screening for CRE at ICU admission; daily chlorhexidine bathing in ICUs; terminal cleaning audited with fluorescent markers |
Continuous |
IPC team |
|
Education and feedback |
Induction and quarterly sessions for residents and nurses; monthly unit-level HAI and antibiotic-use reports |
Monthly |
AMS and IPC teams |
Antibiotics were grouped into Access, Watch and Reserve categories using the WHO AWaRe classification.²⁹
The preintervention period had 18,950 admissions and 130,272 patient-days. The postintervention period had 19,438 admissions and 130,852 patient-days. The two populations were well balanced: no baseline characteristic had an SMD above 0.10 (Table 2). With these large samples, small differences in age, Charlson index and ICU admission reached statistical significance. Mean length of stay was 6.9 days before and 6.7 days after the intervention.
Table 2. Baseline characteristics and exposure denominators
|
Characteristic |
Preintervention (n = 18,950) |
Postintervention (n = 19,438) |
SMD |
|
Age, mean ± SD, y |
52.7 ± 16.7 |
53.5 ± 16.8 |
0.044 |
|
Male, No. (%) |
11,070 (58.4) |
11,387 (58.6) |
0.003 |
|
Charlson Comorbidity Index, median (IQR) |
3 (2–4) |
3 (2–4) |
0.055 |
|
ICU admission, No. (%) |
3,021 (15.9) |
3,260 (16.8) |
0.022 |
|
Surgical admission, No. (%) |
6,488 (34.2) |
6,778 (34.9) |
0.013 |
|
Diabetes mellitus, No. (%) |
4,430 (23.4) |
4,595 (23.6) |
0.006 |
|
Chronic kidney disease, No. (%) |
1,582 (8.3) |
1,734 (8.9) |
0.020 |
|
Active malignancy, No. (%) |
1,286 (6.8) |
1,348 (6.9) |
0.006 |
|
Transferred from another facility, No. (%) |
3,579 (18.9) |
3,727 (19.2) |
0.007 |
|
Patient-days |
130,272 |
130,852 |
— |
|
ICU patient-days (% of total) |
19,157 (14.7) |
19,078 (14.6) |
— |
|
Clean and clean-contaminated procedures |
6,464 |
6,566 |
— |
|
Central line utilisation ratio |
0.43 |
0.40 |
— |
|
Urinary catheter utilisation ratio |
0.63 |
0.56 |
— |
|
Ventilator utilisation ratio |
0.32 |
0.30 |
— |
Abbreviation: SMD, standardised mean difference. Utilisation ratios are device-days divided by ICU patient-days.
All process measures improved (Table 3). Hand hygiene compliance rose from 45.8% to 78.4%, an absolute gain of 32.5 percentage points. Compliance with the 3 device bundles roughly doubled to 83%–87%. Surgical prophylaxis continued beyond 24 hours fell from 65.0% to 22.5%. The AMS team made 1735 PAF recommendations, and prescribers accepted 1305 (75.2%; 95% CI, 73.1%–77.2%).
Table 3. Process measures before and after the intervention
|
Measure |
Preintervention, No./total (%) |
Postintervention, No./total (%) |
RR (95% CI) |
P value |
|
Hand hygiene compliance (observed opportunities) |
2241/4890 (45.8) |
4175/5328 (78.4) |
1.71 (1.65–1.77) |
< .001 |
|
CLABSI insertion and maintenance bundle |
122/312 (39.1) |
287/344 (83.4) |
2.13 (1.84–2.47) |
< .001 |
|
VAP prevention bundle |
135/296 (45.6) |
276/330 (83.6) |
1.83 (1.61–2.10) |
< .001 |
|
CAUTI prevention bundle |
121/288 (42.0) |
277/318 (87.1) |
2.07 (1.80–2.39) |
< .001 |
|
Blood culture before first antibiotic dose |
432/820 (52.7) |
688/864 (79.6) |
1.51 (1.40–1.63) |
< .001 |
|
Empirical therapy matching the hospital policy |
257/560 (45.9) |
466/602 (77.4) |
1.69 (1.53–1.86) |
< .001 |
|
48–72 h time-out documented |
50/600 (8.3) |
462/640 (72.2) |
8.66 (6.61–11.34) |
< .001 |
|
Surgical prophylaxis beyond 24 h |
416/640 (65.0) |
150/668 (22.5) |
0.35 (0.30–0.40) |
< .001 |
The study recorded 778 HAIs before and 456 after the intervention. Overall incidence fell from 5.97 to 3.48 per 1000 patient-days, a 41.6% reduction (IRR, 0.58; 95% CI, 0.52–0.66; P < .001) (Table 4).
The preintervention trend was flat (monthly IRR, 1.008; 95% CI, 0.986–1.030; P = .48). The intervention was followed by an immediate level drop (IRR, 0.63; 95% CI, 0.48–0.82; P < .001). The postintervention slope also declined, but not significantly (monthly IRR, 0.971; 95% CI, 0.937–1.006; P = .11) (Table 7; Figure 2). Taking level and slope together, incidence at month 12 was an estimated 56% below the projected counterfactual. After adjustment for case mix, the level change was unchanged (IRR, 0.64; 95% CI, 0.49–0.84; P = .001).
Figure 2 · approximate working data, 27 months (Jan 2024–Mar 2026); fit from the quasi-Poisson segmented model, plotted for non-monsoon months
By month 12 after the intervention, the fitted rate (2.93) was 56% below the projected preintervention trend (6.65).
Every HAI type declined (Table 4). CLABSI fell by 48.3% (IRR, 0.52; 95% CI, 0.36–0.74), CAUTI by 41.0% (IRR, 0.59; 95% CI, 0.40–0.88) and VAP by 36.8% (IRR, 0.63; 95% CI, 0.47–0.84). SSI fell from 4.6% to 2.7% of procedures (RR, 0.59; 95% CI, 0.49–0.71). In ITS analysis, device-associated HAI per 1000 ICU patient-days showed a level drop of 49% (IRR, 0.51; 95% CI, 0.33–0.77; P = .002). Device utilisation also fell, most for urinary catheters (0.63 to 0.56).
Table 4. Healthcare-associated infection outcomes
|
Outcome |
Preintervention, events/exposure |
Preintervention rate |
Postintervention, events/exposure |
Postintervention rate |
IRR or RR (95% CI) |
P value |
|
Overall HAI, per 1000 patient-days |
778/130,272 |
5.97 |
456/130,852 |
3.48 |
0.58 (0.52–0.66) |
< .001 |
|
Device-associated HAI, per 1000 device-days |
288/26,294 |
10.95 |
155/23,953 |
6.47 |
0.59 (0.49–0.72) |
< .001 |
|
CLABSI, per 1000 central line–days |
93/8179 |
11.37 |
45/7652 |
5.88 |
0.52 (0.36–0.74) |
< .001 |
|
CAUTI, per 1000 urinary catheter–days |
71/12,049 |
5.89 |
37/10,648 |
3.47 |
0.59 (0.40–0.88) |
.009 |
|
VAP, per 1000 ventilator-days |
124/6066 |
20.44 |
73/5653 |
12.91 |
0.63 (0.47–0.84) |
.002 |
|
Other HAIs, per 1000 patient-days |
190/130,272 |
1.46 |
121/130,852 |
0.92 |
0.63 (0.50–0.80) |
< .001 |
|
SSI, per 100 procedures |
300/6464 |
4.6 |
180/6566 |
2.7 |
0.59 (0.49–0.71)ᵃ |
< .001 |
Abbreviations: CAUTI, catheter-associated urinary tract infection; CLABSI, central line–associated bloodstream infection; IRR, incidence rate ratio; RR, risk ratio; SSI, surgical site infection; VAP, ventilator-associated pneumonia. ᵃ Risk ratio. All other ratios are IRRs with exact conditional P values.
Mean monthly antimicrobial use fell from 920.2 to 761.6 DOT per 1000 patient-days, a 17.2% reduction (Table 5). Carbapenem use fell by 35.6%, polymyxin use by 31.3% and Reserve-group use by 36.3%. The share of Access antibiotics rose from 28.5% to 38.7% of all DOT.
In ITS analysis, carbapenem use dropped by 39.1 DOT per 1000 patient-days at the start of the postintervention period (95% CI, −42.7 to −35.6). It then fell by a further 2.2 DOT per month (95% CI, −2.7 to −1.7), reversing a rising preintervention trend (Table 7; Figure 3).
Table 5. Antimicrobial consumption, DOT per 1000 patient-days
|
Antimicrobial group |
Preintervention, mean ± SD |
Postintervention, mean ± SD |
Absolute change |
Relative change, % |
P value |
|
All systemic antibacterials |
920.2 ± 28.4 |
761.6 ± 22.8 |
−158.6 |
−17.2 |
< .001 |
|
Third-generation cephalosporins |
190.5 ± 6.2 |
156.3 ± 6.2 |
−34.2 |
−18.0 |
< .001 |
|
Carbapenems |
121.6 ± 4.1 |
78.3 ± 5.9 |
−43.3 |
−35.6 |
< .001 |
|
Polymyxins |
32.3 ± 1.3 |
22.2 ± 1.1 |
−10.1 |
−31.3 |
< .001 |
|
WHO AWaRe Reserve group |
55.2 ± 2.4 |
35.1 ± 1.6 |
−20.1 |
−36.3 |
< .001 |
|
WHO AWaRe Access group |
261.7 ± 8.6 |
295.0 ± 9.1 |
+33.2 |
+12.7 |
< .001 |
|
Access share of total DOT, % |
28.5 |
38.7 |
+10.3 points |
RR 1.36 (1.35–1.38) |
< .001 |
Abbreviation: DOT, days of therapy. Means are of 12 monthly rates per period, compared by t test.
Figure 3 · approximate working data, 27 months (Jan 2024–Mar 2026); fit from segmented linear regression with Newey-West errors, plotted for non-monsoon months
The projected line extends a rising preintervention trend, so it overstates the counterfactual if that rise would have levelled off. The observed fall from 121.6 to 78.3 DOT per 1000 patient-days does not depend on this projection.
Carbapenem resistance fell in all 3 priority gram-negative pathogens (Table 6). It fell in K pneumoniae from 67.0% to 59.7% (RR, 0.89; 95% CI, 0.80–0.99; P = .03), in A baumannii from 88.7% to 80.7% (RR, 0.91; 95% CI, 0.85–0.98; P = .01) and in P aeruginosa from 42.2% to 29.3% (RR, 0.69; 95% CI, 0.53–0.91; P = .008). 3GC resistance in E coli fell from 81.0% to 72.7% (P = .009).
The declines in carbapenem-resistant E coli, MRSA and VRE were not statistically significant. Hospital-onset incidence fell by 31% for CRE, 32% for CRAB and 43% for MRSA.
The nonequivalent dependent variable did not change. 3GC resistance in community-onset outpatient E coli was 58.7% before and 58.8% after the intervention (RR, 1.00; 95% CI, 0.94–1.07; P = .98).
Table 6. Resistance proportions and MDRO incidence in hospital-onset clinical isolates
|
Organism and resistance marker |
Preintervention, resistant/tested (%) |
Postintervention, resistant/tested (%) |
RR or IRR (95% CI) |
P value |
|
K pneumoniae, carbapenem |
280/418 (67.0) |
222/372 (59.7) |
0.89 (0.80–0.99) |
.03 |
|
K pneumoniae, 3GC |
355/418 (84.9) |
304/372 (81.7) |
0.96 (0.90–1.02) |
.23 |
|
E coli, carbapenem |
68/352 (19.3) |
51/344 (14.8) |
0.77 (0.55–1.07) |
.12 |
|
E coli, 3GC |
285/352 (81.0) |
250/344 (72.7) |
0.90 (0.83–0.97) |
.009 |
|
A baumannii, carbapenem |
258/291 (88.7) |
197/244 (80.7) |
0.91 (0.85–0.98) |
.01 |
|
P aeruginosa, carbapenem |
86/204 (42.2) |
55/188 (29.3) |
0.69 (0.53–0.91) |
.008 |
|
S aureus, methicillin (MRSA) |
87/168 (51.8) |
67/152 (44.1) |
0.85 (0.68–1.07) |
.17 |
|
Enterococcus spp, vancomycin (VRE) |
12/121 (9.9) |
12/126 (9.5) |
0.96 (0.45–2.05) |
.92 |
|
CRE incidence, per 1000 patient-days |
162 (1.24) |
113 (0.86) |
0.69 (0.55–0.88)ᵃ |
.003 |
|
CRAB incidence, per 1000 patient-days |
134 (1.03) |
92 (0.70) |
0.68 (0.52–0.89)ᵃ |
.005 |
|
MRSA incidence, per 1000 patient-days |
66 (0.51) |
38 (0.29) |
0.57 (0.38–0.85)ᵃ |
.006 |
|
Control outcome: community-onset OPD urine E coli, 3GC |
728/1240 (58.7) |
745/1268 (58.8) |
1.00 (0.94–1.07) |
.98 |
Abbreviations: 3GC, third-generation cephalosporin; CRAB, carbapenem-resistant A baumannii; CRE, carbapenem-resistant Enterobacterales; MRSA, methicillin-resistant S aureus; OPD, outpatient department; VRE, vancomycin-resistant enterococci. ᵃ IRR. Incidence rows show No. of new isolates (rate).
Table 7 gives all segmented-regression estimates. The models showed no material overdispersion (Pearson χ²/df, 0.86–1.02) and no significant residual autocorrelation at lag 6.
Table 7. Segmented regression estimates
|
Outcome (model) |
Preintervention trend, per month (95% CI) |
Level change (95% CI) |
Slope change, per month (95% CI) |
Durbin-Watson |
|
Overall HAI per 1000 patient-days (quasi-Poisson), IRR |
1.008 (0.986–1.030) |
0.63 (0.48–0.82); P < .001 |
0.971 (0.937–1.006); P = .11 |
2.76 |
|
Overall HAI, adjusted for case mixᵃ, IRR |
1.003 (0.983–1.024) |
0.64 (0.49–0.84); P = .001 |
0.974 (0.940–1.009); P = .15 |
2.87 |
|
Device-associated HAI per 1000 ICU patient-days, IRR |
1.013 (0.980–1.046) |
0.51 (0.33–0.77); P = .002 |
0.982 (0.928–1.038); P = .51 |
2.68 |
|
Total DOT per 1000 patient-days (linear, Newey-West), β |
+6.1 (3.7 to 8.4) |
−181.0 (−205.8 to −156.2); P < .001 |
−10.5 (−13.7 to −7.4); P < .001 |
1.86 |
|
Carbapenem DOT per 1000 patient-days (linear, Newey-West), β |
+0.7 (0.3 to 1.1) |
−39.1 (−42.7 to −35.6); P < .001 |
−2.2 (−2.7 to −1.7); P < .001 |
1.59 |
All models include a monsoon-season term (July–September). ᵃ Adjusted for monthly mean Charlson Comorbidity Index and ICU share of patient-days.
The results held in both sensitivity analyses. Negative binomial regression gave the same level change (IRR, 0.63; 95% CI, 0.48–0.82). The estimated dispersion was close to 0, which supports the Poisson model. Counting the implementation months as postintervention made the effect smaller but it stayed significant (IRR, 0.77; 95% CI, 0.63–0.95; P = .01). This fits a phased roll-out during January to March 2025.
In-hospital mortality was 6.7% (1273/18,950) before and 6.4% (1250/19,438) after the intervention (RR, 0.96; 95% CI, 0.89–1.03; P = .26). Median length of stay was unchanged at 7 days (IQR, 4–10).
Principal findings
In this tertiary teaching hospital, a microbiologist-led AMS and IPC programme was associated with a 42% lower overall HAI incidence. The ITS level change was 37% and survived adjustment for case mix. Every HAI type declined, and CLABSI showed the largest relative fall. Carbapenem use fell by over a third, and the share of Access antibiotics rose by 10 percentage points. Carbapenem resistance fell in K pneumoniae, A baumannii and P aeruginosa, and hospital-onset CRE, CRAB and MRSA incidence fell by 31% to 43%. Mortality did not rise. Resistance in community-onset E coli, which the intervention should not affect, did not change. This makes a hospital-wide secular trend or a laboratory artefact an unlikely explanation.
Mechanisms
The two arms of the intervention probably worked through different but reinforcing pathways. The IPC arm acted directly on infection risk. Bundle compliance roughly doubled, hand hygiene compliance rose by 33 points, and daily review of device need reduced catheter and central line use. The device utilisation ratios fell, so part of the effect on device-associated infection came simply from fewer device-days. We therefore treated device use as a mediator and did not adjust for it.
The AMS arm reduced antibiotic selection pressure, mainly from carbapenems and polymyxins. Lower carbapenem exposure reduces the selective advantage of carbapenemase-producing strains. Contact precautions, admission screening and chlorhexidine bathing then limit their onward spread. The meta-analysis by Baur et al found that stewardship reduced MDR gram-negative colonisation and infection most when it was paired with infection control.¹² The fall in hospital-onset CRE and CRAB incidence (31% and 32%), larger than the fall in resistance proportions (7 to 8 points), fits this joint mechanism: fewer resistant organisms were acquired, and fewer patients were infected at all.
Comparison with previous studies
Our baseline device-associated infection rates were high compared with Indian network data. CLABSI was 11.4 and VAP 20.4 per 1000 device-days, against pooled INICC rates of 5.1 and 9.4 in Indian adult and paediatric ICUs.⁴ Higher baselines leave more room for improvement, which may partly explain the size of our effect. The 48% CLABSI reduction is close to the effect of checklist-based programmes in high-income ICUs.¹⁴ The 41% CAUTI reduction is larger than in a US national programme, where reductions were seen in non-ICU units but not in ICUs.³⁷
Our 17% fall in total consumption is close to the 19% pooled reduction reported in a meta-analysis of hospital stewardship programmes.³⁸ The 75% acceptance of PAF recommendations is higher than in the earlier stewardship programme at an Indian tertiary centre, which reported 54% acceptance.¹⁶ Daily presence of a clinical microbiologist on the wards, rather than written advice alone, may have helped.
A meta-analysis of Asian programmes found that stewardship reduced antibiotic use but improved HAI rates only when infection control or hand hygiene was part of the programme.¹⁵ Our data support that conclusion. They also add resistance and MDRO outcomes, which earlier Indian evaluations have rarely reported.¹⁶˒¹⁷
Resistance outcomes
The resistance changes were modest and organism-specific, as expected over 12 months. A baumannii stayed above 80% carbapenem resistant after the intervention. This matches the hyperendemic levels in Indian ICUs, where it is one of the hardest pathogens to control by stewardship alone.⁵˒⁶ The changes in carbapenem-resistant E coli, MRSA and VRE did not reach significance. The study had limited power for these smaller isolate groups, and these results should not be read as evidence of no effect. Without molecular typing we cannot tell whether the declines reflect less clonal transmission or less selection; both mechanisms are plausible.³⁹
Implications for practice
Most Indian hospitals do not have infectious diseases physicians or specialist pharmacists. This programme was led by a clinical microbiologist and used existing staff, printed checklists and the hospital antibiogram. Most components cost little: handrub, checklists, audits and feedback. The model fits the ICMR AMSP guideline and the hospital-level actions in the National Action Plan on AMR 2.0.¹⁸˒¹⁹ The early level drop and the continuing downward slope in carbapenem use suggest that the changes held over the year. Sustainment beyond 12 months still needs to be shown.
Hospitals planning similar programmes should run AMS and IPC together, not in sequence. They should also measure process, outcome and balancing indicators from the start.⁴⁰
Strengths and limitations
The study has several strengths. HAI surveillance used the same definitions and team in both periods, and data were collected prospectively. Twelve monthly points on each side of the intervention allowed segmented regression. A nonequivalent dependent variable tested for secular confounding. Process measures documented that the intervention was delivered, and balancing measures checked for harm.
The study also has important limitations:
A structured, microbiologist-led programme combining antimicrobial stewardship with infection prevention and control was associated, within 12 months, with 42% fewer HAIs, 36% less carbapenem use and lower carbapenem resistance in the main gram-negative pathogens. It did not increase mortality. The programme used existing staff and low-cost tools, so Indian tertiary hospitals implementing the National Action Plan on AMR 2.0 could adopt it.
Priorities for future research: