Introduction
Material and methods
Spanish ICU data
General
|
Frequency
| |
Number of admissions | 109,216 | |
Number of admission-days | 813,739 | |
Number of ICUs | 159 | |
Number of nosocomial bacteremia during ICU stay | 5,498 (5.03%) | |
Number of deaths without NB during ICU stay | 12,678 (11.61%) | |
Number of discharges without NB from ICU | 90,142 (82.54%) | |
Number of administrative censored admissions | 898 (0.82%) | |
Overall risk of nosocomial bacteremia (censored excluded) | 5.08% | |
Overall rate of nosocomial bacteremia | 6.75 / 1,000 admission-days | |
Risk factors
|
Frequency (%)
|
Frequency (%)
|
(patient level)
|
(ICU level)
| |
ICU / hospital level covariates
| ||
Number of beds in ICU: | ||
0 to 10 (reference) | 30,389 (27.82) | 61 (38.36) |
11 to 20 | 46,524 (42.60) | 71 (44.65) |
21 to 30 | 21,668 (19.84) | 19 (11.95) |
31 to 40 | 7,673 (7.03) | 5 (3.14) |
> 40 | 2,962 (2.71) | 3 (1.89) |
Number of beds in hospital: | ||
0 to 500 (reference) | 52,426 (48.00) | 94 (59.12) |
501 to 1,000 | 8,259 (7.56) | 10 (6.29) |
> 1,000 | 48,531 (44.44) | 55 (34.59) |
Type of hospital: | ||
Private | 6,541 (5.99) | 12 (7.55) |
Public | 102,675 (94.01) | 147 (92.45) |
Type of ICU: | ||
Polyvalent (reference) | 96,478 (88.34) | 138 (86.79) |
Medical | 3,895 (3.57) | 5 (3.14) |
Surgery | 3,103 (2.84) | 3 (1.89) |
Coronary | 525 (0.48) | 2 (1.26) |
Traumatology | 3,273 (3.00) | 6 (3.77) |
Post-surgery cardiology | 1,834 (1.68) | 4 (2.52) |
Burn | 108 (0.10) | 1 (0.63) |
University+teaching hospital (reference) | 67,917 (62.19) | 86 (54.09) |
Teaching hospital (no university) | 30,089 (27.55) | 52 (32.70) |
Hospital without teaching/university | 11,210 (10.26) | 21 (13.21) |
Calendar year of admission
| ||
2006 (reference) | 14,318 (13.11) | |
2007 | 17,819 (16.32) | |
2008 | 21,559 (19.74) | |
2009 | 25,660 (23.49) | |
2010+ | 29,860 (27.34) | |
Patient level covariates
| ||
APACHE II score: | ||
0 to 10 (reference) | 40,353 (36.95) | |
11 to 20 | 44,654 (40.89) | |
21 to 30 | 19,191 (17.57) | |
> 30 | 5,018 (4.59) | |
Age (years): | ||
0 to 40 | 54,477 (49.88) | |
41 to 60 | 12,931 (11.84) | |
61 to 80 (reference) | 29,371 (26.89) | |
> 80 | 12,437 (11.39) | |
Days in hospital before ICU admission: | ||
0 to 3 (reference) | 87,208 (79.85) | |
4 to 6 | 5,864 (5.37) | |
7 to 10 | 4,962 (4.54) | |
> 10 | 11,182 (10.24) | |
Type of diagnosis: | ||
Cardiovascular (reference) | 54,374 (49.79) | |
Respiratory | 15,243 (13.96) | |
Gastrointestinal | 14,626 (13.39) | |
Central nervous system | 17,567 (16.08) | |
Other diagnoses | 7,406 (6.78) | |
Antibiotic treatment 48 h before and/or after ICU admission | 23,178 (21.22) | |
Gender (male) | 71,223 (65.21) | |
Origin: community (reference) | 54,996 (50.36) | |
Origin: hospital/ICU | 54,220 (49.64) | |
Trauma | 8,927 (8.17) |
Shared frailty models for competing risks
Model 1: etiologic model (rate-based)
Model 2: predictive model (risk-based)
Results
Baseline hazard rates and cumulative incidence function
Heterogeneity across ICUs
Risk factors | Event-specific analysis | Subdistribution analysis | ||
---|---|---|---|---|
Nosocomial bacteremia (NB)
|
Death without NB in ICU
|
Discharge without NB
|
Nosocomial bacteremia
| |
HR (95% CI) | HR (95% CI) | HR (95% CI) | Subdistribution HR (95% CI) | |
ICU / hospital-level covariates
| ||||
Number of beds in ICU: | ||||
11–20 vs 0–10 | 1.18 (0.96–1.46) | 1.01 (0.87–1.17) | 1.43 (1.23–1.67) | 1.34 (0.99–1.81) |
21–30 vs 0–10 | 1.31 (0.97–1.77) | 0.86 (0.69–1.08) | 1.15 (0.89–1.47) | 2.17 (1.42–3.30) |
31–40 vs 0–10 | 1.56 (0.98–2.50) | 0.81 (0.57–1.14) | 1.83 (1.22–2.74) | 1.85 (0.95–3.62) |
> 40 vs 0–10 | 1.09 (0.55–2.18) | 1.00 (0.60–1.65) | 1.37 (0.76–2.48) | 1.37 (0.50–3.72) |
Number of beds in hospital: | ||||
> 1000 vs 0–500 | 1.30 (0.85–1.99) | 0.94 (0.70–1.27) | 1.30 (0.90–1.87) | 1.32 (0.71–2.47) |
501–1000 vs 0–500 | 1.11 (0.91–1.36) | 1.04 (0.90–1.21) | 1.13 (0.95–1.34) | 1.16 (0.87–1.53) |
Teaching hospital (only) vs university+teaching | 1.00 (0.82–1.22) | 1.03 (0.89–1.20) | 1.87 (1.61–2.18) | 0.94 (0.71–1.26) |
No teaching hospital vs university+teaching | 0.71 (0.51–0.97) | 1.08 (0.86–1.34) | 2.42 (1.92–3.03) | 0.59 (0.38–0.91) |
Type of hospital (private vs public) | 0.98 (0.70–1.39) | 0.82 (0.64–1.05) | 0.80 (0.61–1.05) | 1.18 (0.76–1.85) |
Type of ICU:
| ||||
Medical vs polyvalent | 0.71 (0.44–1.14) | 1.07 (0.78–1.47) | 1.44 (0.99–2.10) | 0.57 (0.30–1.08) |
Surgery vs polyvalent | 1.07 (0.62–1.85) | 0.60 (0.40–0.91) | 1.20 (0.73–1.96) | 1.13 (0.51–2.50) |
Coronary vs polyvalent | 0.71 (0.28–1.82) | 1.27 (0.64–2.51) | 0.83 (0.44–1.57) | 0.73 (0.24–2.28) |
Traumatology vs polyvalent | 1.20 (0.79–1.82) | 1.03 (0.76–1.40) | 1.28 (0.90–1.81) | 1.15 (0.64–2.05) |
Post-surgery cardiology vs polyvalent | 1.47 (0.85–2.54) | 0.71 (0.46–1.08) | 1.95 (1.27–3.01) | 1.39 (0.63–3.05) |
Burn vs polyvalent | 0.48 (0.16–1.42) | 0.70 (0.31–1.62) | 0.70 (0.31–1.59) | 1.20 (0.30–4.79) |
Calendar year of admission
| ||||
2007 vs 2006 | 1.02 (0.92–1.13) | 1.01 (0.94–1.08) | 0.95 (0.92–0.97) | 1.02 (0.93–1.13) |
2008 vs 2006 | 0.98 (0.89–1.08) | 1.09 (1.02–1.16) | 0.93 (0.91–0.95) | 1.03 (0.93–1.14) |
2009 vs 2006 | 1.09 (0.99–1.19) | 1.02 (0.96–1.09) | 0.86 (0.83–0.88) | 1.19 (1.09–1.31) |
2010+ vs 2006 | 0.84 (0.76–0.93) | 1.08 (1.01–1.15) | 0.94 (0.91–0.96) | 0.83 (0.75–0.91) |
Patient level covariates
| ||||
APACHE II score: | ||||
11–20 vs 0–10 | 1.30 (1.20–1.42) | 1.99 (1.85–2.14) | 0.53 (0.52–0.54) | 2.81 (2.59–3.06) |
21–30 vs 0–10 | 1.38 (1.26–1.50) | 4.11 (3.83–4.41) | 0.28 (0.27–0.28) | 4.54 (4.15–4.96) |
> 31 vs 0–10 | 1.54 (1.37–1.73) | 6.65 (6.15–7.19) | 0.17 (0.17–0.18) | 5.78 (5.14–6.49) |
Age (years): | ||||
0–40 vs 61–80 | 1.08 (0.99–1.18) | 0.59 (0.55–0.64) | 1.05 (1.03–1.07) | 1.12 (1.02–1.22) |
40–60 vs 61–80 | 1.09 (1.02–1.16) | 0.79 (0.76–0.83) | 0.98 (0.97–1.00) | 1.18 (1.11–1.26) |
> 80 vs 61–80 | 0.77 (0.68–0.86) | 1.65 (1.57–1.73) | 1.18 (1.15–1.21) | 0.51 (0.45–0.57) |
Days in hospital before ICU admission:
| ||||
4–6 vs 0–3 | 1.12 (1.00–1.25) | 1.09 (1.01–1.17) | 0.86 (0.83–0.89) | 1.27 (1.13–1.42) |
6–10 vs 0–3 | 1.14 (1.01–1.28) | 1.18 (1.10–1.27) | 0.85 (0.82–0.87) | 1.33 (1.18–1.49) |
> 10 vs 0–3 | 1.17 (1.08–1.27) | 1.22 (1.16–1.29) | 0.78 (0.76–0.80) | 1.43 (1.32–1.55) |
Type of diagnosis:
| ||||
Respiratory vs cardiovascular | 0.86 (0.80–0.94) | 0.93 (0.88–0.98) | 0.63 (0.62–0.65) | 1.43 (1.32–1.55) |
Gastrointestinal vs cardiovascular | 1.18 (1.09–1.29) | 0.96 (0.91–1.02) | 0.78 (0.76–0.79) | 1.65 (1.51–1.80) |
Central nervous system vs cardiovascular | 0.91 (0.84–0.99) | 1.41 (1.34–1.49) | 0.65 (0.64–0.66) | 1.38 (1.28–1.50) |
Other diagnoses vs cardiovascular | 1.20 (1.08–1.32) | 0.78 (0.71–0.85) | 0.76 (0.74–0.78) | 1.96 (1.78–2.16) |
Antibiotic treatment 48 h before and/or after ICU admission | 0.83 (0.77–0.88) | 1.08 (1.04–1.13) | 0.70 (0.69–0.71) | 1.01 (0.95–1.09) |
Gender | 1.08 (1.01–1.14) | 1.01 (0.97–1.05) | 0.99 (0.98–1.01) | 1.12 (1.06–1.18) |
Origin (hospital/ICU vs community) | 1.05 (0.98–1.12) | 0.98 (0.94–1.02) | 1.00 (0.99–1.02) | 1.09 (1.02–1.17) |
Trauma | 1.13 (1.04–1.24) | 0.67 (0.62–0.72) | 0.64 (0.62–0.66) | 1.81 (1.66–1.98) |
Variance of heterogeneity (without covariates) | 0.26 (SE 0.038) | 0.14 (SE 0.019) | 0.15 (SE 0.017) | 0.64 (SE 0.076) |
Variance of heterogeneity (with ICU-level covariates only) | 0.20 (SE 0.030) | 0.12 (SE 0.015) | 0.09 (SE 0.011) | 0.47 (SE 0.061) |
Variance of heterogeneity (with patient-level covariates only) | 0.25 (SE 0.037) | 0.12 (SE 0.016) | 0.17 (SE 0.025) | 0.56 (SE 0.069) |
Variance of heterogeneity (with all covariates) | 0.19 (SE 0.030) | 0.11 (SE 0.015) | 0.17 (SE 0.024) | 0.40 (SE 0.052) |
Multilevel risk factors: patient level
Multilevel risk factors: ICU level
Discussion
Conclusions
Key messages
-
Discharge from and death in an intensive care unit are competing risks for nosocomial infection.
-
There are factors on the patient as well as on the ICU level influencing the occurrence of nosocomial infections.
-
Analysis of data from large multicenter studies has the potential to improve our understanding of how patient and ICU-level characteristics impact nosocomial infections.
-
A combination of multilevel and competing risk models are necessary to analyze such complex data.
-
We encourage further investigations by using our methodological approaches to evaluate our findings of unexplained heterogeneity. The statistical code is available in Additional file 1.