Early identification of sepsis in hospital inpatients by ward nurses increases 30-day survival

Since 1994 the Mid-Norway Sepsis Study has continuously prospectively registered patients with confirmed BSI identified by growth of one or more microbes from blood culture, combined with clinical evidence of systemic infection [8]. A new occurrence of BSI was recorded for patients if at least 30 days had passed since a previous incident [9]. Clinical information from the patients’ records was recorded retrospectively, after standardized definitions, by a team of trained health care professionals for all patients admitted before 2014 [8].

The intervention was implemented from January to October 2011 in one emergency community hospital, serving around 90,000 citizens, in the Mid-Norway Sepsis Study catchment area. The hospital has 124 beds, approximately 15,700 admissions annually, and has 6 beds in the intensive care unit (ICU) and 5 in the high dependency unit (HDU), equivalent to 12.6 ICU and HDU beds/100,000 citizens. The mean patient load across all staff-registered nurses who are involved in bedside work is 0.24 [10]. The hospital does not have a formal Critical Care Outreach service, but there is daily dialog between the ward physician and the ICU doctors to discuss patients in possible need of transfer to the ICU/HDU. Acute transfers are handled by the ICU doctor on call. The intervention was a bundle that included a flow chart for sepsis identification, treatment and physician response time (see Additional file 1), a SIRS and organ failure triage (SOF-Triage) (Fig. 1) that was used on the wards to re-evaluate for sepsis if clinically indicated. The new flow chart instructed the ward nurse in charge of the sepsis patient to call the ICU doctor on call directly if the ward physician was occupied and prevented from applying within the time limits of the SOF-Triage. Additionally, the bundle included information to all physicians and a new four-hour training course for all nurses and nursing students working on the wards (medical, surgical and gynecological). The training included the pathophysiology, early signs, and treatment of sepsis. The latter included the importance of prompt intravenous (i.v.) fluids and appropriate antibiotics, training in the SOF-Triage, and objective communication about the patients’ status based on observation of vital signs. Six extra training sessions (on the background of the project, interpretation of blood gas analysis, use of the SOF-Triage and flow chart) was given to twelve expert nurses (at least one nurse in each ward), who assisted in the implementation of the intervention.

The pre-intervention group included patients with confirmed BSI in the period from January 2008 to December 2010 as this period had guidelines that mandated equal vigilance towards observation of vital signs as the intervention group (respiratory frequency, heart rate, and temperature should be observed at least every 4 hours per first 24 hours in patients with suspected sepsis). The post-intervention group consisted of patients with BSI admitted after implementation of the intervention, i.e., from Nov. 2011 to Dec. 2013.

The variables were categorized as follows: the observation-rate of each vital sign the first 24 hours after drawing a blood culture (temperature, respiratory frequency or heart rate) was graded as poor (0 or 1 recorded observation), some (2–4 recorded observations) or good observation (≥5 recorded observations); sepsis at admission as BSI without signs of SIRS, sepsis, severe sepsis, or septic shock [6]; sepsis-related Sequential Organ Failure Assessment (SOFA score) [11] at baseline and during follow up was dichotomized as non-severe or severe organ failure (corresponding to infection-related SOFA score >2 in any organ) [8]; infection focus as lungs, urinary tract, and other origin; BSI-type as gram-negative, gram-positive, and mixed microbial BSI (Additional file 2); length of stay (LOS) in the HDU and ICU and survival were counted in days, 0–24 hours was coded as 1 day, >24–48 hours as 2 days and so forth. All patients were censored after 30 days of observation; age (in years) at the time of positive blood culture was treated as a continuous variable and as a categorical variable (<65 years, 65–79 years, and >80 years); McCabe score was used to exclude patients with rapidly fatal illness at admission (death expected within 1 month) [12]; the Charlson weighted Comorbidity Index (CCI) was classified as low (no underlying disease score), medium (score 1–2), or high (score >2) [13]; place of acquisition was categorized as hospital-acquired if a positive blood culture was drawn >48 hours after admission, healthcare-associated (HCA) or community-acquired as defined by Friedman et al. [14], with the exception that HCA infection was defined for patients hospitalized for ≥2 days in the last 30 days (instead of the last 90) before admission, as recommended by Schorr et al. [15]; functional status was categorized as independent if the patient lived at home without any help from the community health service, partly independent if the patient had some help from the community health service, e.g., with medication or wound care, and dependent if the patient lived in a nursing home or lived at home with help in most of the daily life activities [8]; appropriate empiric antibiotic therapy was defined if (1) given intravenously in correct doses within 24 hours after the blood culture and (2) active in vitro against the isolated microbe(s) [16]; intravenous fluid was documented as amount in mL given during the first 24 hours after diagnosis.

A total of 478 BSI patients were admitted in the pre-intervention period and 422 in the post-intervention period. We excluded 19 patients with rapidly fatal illness, 6 (1.26 %) in the pre-intervention group and 13 (3.1 %) in the post-intervention group (p = 0.001). Thus, the pre-intervention group included 472 BSI episodes, and the post-intervention group included 409 BSI episodes. The post-intervention group had a higher proportion of patients with a severe SOFA score and the patients were less in need of help in daily life than in the pre-intervention group. For more details about the groups see Table 1. The ratios of gram-negative, gram-positive and mixed microbial BSI were similar in the two groups as Table 1 shows.

The nurses in the post-intervention group increased their observation frequency of all vital signs (see details in Fig. 2). The post-intervention group had better observations of all vital signs in both patients with and without organ failure (p ≤0.002 for all vital signs).

At 7 days after confirmed BSI, 22 patients (4.6 %) in the pre-intervention group had died compared to 14 (3.4 %) patients in the post-intervention group (p = 0.36). After 30 days 59 (12.5 %) patients in the pre-intervention group had died versus 29 (7.1 %) patients in the post-intervention group (p = 0.035). Table 2 displays the ORs (95 % CIs) for surviving 7 and 30 days after BSI diagnosis in the pre-intervention and post-intervention groups in the different adjustment models.

In additional adjustment analyses for the variables immunosuppression, focus of infection, place of acquisition, and age in categories attenuated the impact of the intervention from 2.7 towards 2.0 (95 % CI 1.2, 3.3), see Additional file 3 for the details.

On sensitivity analyses for the first incidents of sepsis (n = 738) there was a stronger effect on survival in the post-intervention group than if all episodes of BSI (n = 881) were included. The post-intervention group with the first incidents of BSI had an OR of 2.4 (95 % CI 1.0, 5.3) for surviving 7 days and an OR of 3.6 (95 % CI 1.9, 6.9) for surviving 30 days compared to the pre-intervention group. The additional analyses for the first incidents of sepsis are provided in Additional file 4 (baseline characteristics of first-time incidents of BSIs) and Additional file 5 (odds of surviving 7 and 30 days in the post-intervention group). We found no evidence of interaction between survival and age or sex in any of the analysis (p > 0.1 for all).