Frequency of Vital Signs Monitoring and its Association with Mortality among Adults with Severe Sepsis Admitted to a General Medical Ward in Uganda

Introduction Optimal vital signs monitoring of patients with severe sepsis in resource-limited settings may improve outcomes. The objective of this study was to determine the frequency of vital signs monitoring of patients with severe sepsis and its association with mortality in a regional referral hospital in Uganda. Methods We reviewed medical records of patients admitted to Mbarara Regional Referral Hospital in Southwestern Uganda with severe sepsis defined by the presence of infection plus ≥2 of the systemic inflammatory response syndrome criteria, and ≥1 organ dysfunction (altered mental state, hypotension, jaundice, or thrombocytopenia). We recorded frequency of vital signs monitoring in addition to socio-demographic, clinical, and outcome data. We analyzed the data using logistic regression. Results We identified 202 patients with severe sepsis. The median age was 35 years (IQR, 25–47) and 98 (48%) were female. HIV infection and anemia was present in 115 (57%) and 83 (41%) patients respectively. There were 67 (33%) in-hospital deaths. The median monitoring frequency per day was 1.1 (IQR 0.9–1.5) for blood pressure, 1.0 (IQR, 0.8–1.3) for temperature and pulse, and 0.5 (IQR, 0.3–1.0) for respiratory rate. The frequency of vital signs monitoring decreased during the course of hospitalization. Patients who died had a higher frequency of vital signs monitoring (p<0.05). The admission respiratory rate was associated with both frequency of monitoring (coefficient of linear regression 0.6, 95% CI 0.5–0.8, p<0.001) and mortality (AOR 2.5, 95% CI 1.3–5.3, p = 0.01). Other predictors of mortality included severity of illness, HIV infection, and anemia (p<0.05). Conclusions More research is needed to determine the optimal frequency of vital signs monitoring for severely septic patients in resource-limited settings such as Uganda.


Introduction
Severe sepsis is a serious illness that occurs when infection leads to organ failure and frequently causes death, especially if it is not rapidly recognized and treated [1]. Application of recent guidelines that emphasize early goal-directed resuscitation, prompt antibiotic administration, oxygen therapy, inotropic medications, and blood transfusion can reduce mortality in resource-rich settings [2,3]. Modifications have been recommended for resource-limited settings, such as sub-Saharan Africa (SSA), which shoulder the largest burden of sepsis, but their applicability has been questioned [4][5][6]. Close monitoring of patients with severe sepsis may improve outcomes by identifying patients at risk of deterioration which can lead to appropriate interventions [1,3,7,8].
The surviving sepsis campaign's international guidelines for management of sepsis do not make specific recommendations on the frequency of vital signs monitoring [1]. However, in resourcerich settings, management of severe sepsis usually takes place in emergency departments or intensive care units where continuous monitoring is possible [9]. The initial management should be goaldirected and iterative, hence frequent monitoring is necessary to meet pre-set targets [1,3]. In contrast, in resource-limited settings, severely septic patients are typically managed on general wards [7,[10][11][12]. For patients with severe sepsis, the Integrated Management of Adolescent and Adult Illness (IMAI) guidelines for district hospitals in resource-limited settings created by the World Health Organization (WHO) recommend administration of antibiotics in the first hour along with monitoring blood pressure (BP), pulse, and respiratory rate every 30 minutes to hourly in the first 6 hours of admission, and hourly to every 2 hours thereafter for the next 24 hours [13,14]. Additionally, temperature should be measured every 6 hours. After the first 24 hours, vital signs monitoring in the post-resuscitation period is recommended 1 to 3 times daily.
The current frequency of vital sign monitoring of patients with severe sepsis and the applicability of the IMAI guidelines in resource-limited settings such as government referral hospitals in SSA is unknown. Therefore, our objective was to determine the frequency of vital signs monitoring and its association with mortality for patients with severe sepsis managed on the general ward of a regional referral hospital in Uganda.

Materials and Methods
This was an observational cohort study using medical records from the adult medical ward of Mbarara Regional Referral Hospital (MRRH) in Southwestern Uganda. At MRRH, patients with severe sepsis are initially resuscitated in an emergency room before admission to the ward. Initial diagnoses are made by medical residents and then reviewed by faculty physicians. The facilities available at MRRH have been previously described, but in addition, in September 2012, a new emergency room was opened where patients could stay overnight after admission [11]. This study was carried out in July 2013 and was approved by the Institutional Review Board of Mbarara University of Science and Technology. Since this was a retrospective study using data obtained from hospital charts, the institutional review board waived the need for written informed consent from the participants.
We reviewed admission records from July, October, and November of 2012, and January and February of 2013. We did not review records from holiday periods or times when clinical studies were being conducted on the ward. All patients admitted with a diagnosis of infection during the study period were identified from the hospital admission register. Their medical charts were then screened to identify patients with severe sepsis. Severe sepsis was defined as the presence of infection plus $2 systemic inflammatory response syndrome criteria (pulse $90 beats/minute; respiratory rate $20 cycles/minute; temperature $38uC or #36uC; and white blood cell concentration $12,000 cells/cc or ,4,000 cells/cc or .10% band forms), and at least 1 organ dysfunction (altered mental state, systolic blood pressure [SBP] ,90 or diastolic BP [DBP] ,60 mmHg, jaundice, or platelet count ,100,000 cells/cc).

Measurements
The primary predictor was the frequency of vital signs monitoring including BP, temperature, pulse, and respiratory rate. We abstracted all vital signs recordings available for the duration of each admission together with their associated date and time. We obtained clinical data including diagnosis, antibiotics and intravenous fluids prescribed, along with laboratory data including complete blood counts and HIV serostatus. Anemia was defined as a hemoglobin of ,12 g/dL for men and ,11.5 g/dL for women. Additional information obtained included sex, age, occupation, district of residence, and whether the patient was admitted from the community or from another hospital or clinic. We determined the severity of illness by the total number of dysfunctional organs and a CRB-65 score [11,15]. The CRB-65 score (range 0-4) was determined by the aggregate of points accrued for confusion, respiratory rate $20, systolic BP,90 mmHg or diastolic BP#60 mmHg, and age $65. The primary outcome was inhospital mortality. Other outcomes included self-discharge, referral to a higher level of care facility, readmission, and length of hospital stay.

Analysis
The average frequency of vital signs monitoring was calculated as the total count for each recording of vital signs divided by days in hospital. The association of monitoring frequency with mortality was determined using logistic regression (Stata 12, StataCorp, College Station, Texas, USA). Statistical significance was set at p,0.05. In multivariable models we adjusted for severity of illness as measured by a CRB-65 score, age, sex, anemia, and HIV status. We also assessed the association of post admission changes in blood pressure with mortality. A sample size of 200 patients was required to detect a reduction in mortality of 13% per unit increase in monitoring frequency given an expected mortality rate among severely septic patients of 33.8%, an alpha of 0.05, a beta of 0.2, and a penalty of 0.3 for including other predictors in the multivariable model [11,16].

Frequency of vital signs monitoring and association with mortality
In the first 24 hours of admission, 193 (96%) patients had at least 1 BP measurement, and 153 (76%) had 2 BP measurements. However, only 29 (14%) had 3 BP measurements within the first 24 hours of admission. The median monitoring frequency per day of hospital stay was 1.1 (IQR 0.9-1.5) for BP, 1.0 (IQR, 0.8-1.3) for temperature and pulse, and 0.5 (IQR, 0.3-1.0) for respiratory rate. In both the unadjusted and multivariable-adjusted analyses, an increase in monitoring frequency was associated with higher odds of mortality regardless of the initial severity of illness (Table 2). In the unadjusted analysis, a unit increase in BP monitoring frequency was associated with a 2-fold increase in the odds of mortality (OR 2.0, 95% CI 1.2-3.2, p = 0.004). This association persisted after adjusting for HIV status, anemia, and severity of illness (AOR 2.5, 95% CI 1.4-4.5, p = 0.001).
The frequency of BP monitoring was slightly higher (median 1.3 [IQR, 1.0-2.0]) for patients with .1 dysfunctional organs than for patients with 1 dysfunctional organ (median 1.1 [IQR, 0.9-1.5], p = 0.06). However, the median monitoring frequency for temperature, pulse, and respiratory rate did not vary according to the number of dysfunctional organs, and the frequency of monitoring for all vital signs did not vary according to CRB-65 score, focus of infection, or site of organ dysfunction. The admission respiratory rate did predict the frequency of vital signs monitoring with each unit increase in respiratory rate associated with a 0.6 unit increase in BP monitoring per day (coefficient of linear regression 0.6, 95% CI 0.5-0.8, p,0.001). The frequency of vital signs monitoring decreased throughout the duration of hospitalization ( Figure 2).

Post admission change in blood pressure and mortality
Repeat BP measurements on the day of admission were available for 147 patients. There was a decrease in MAP$10 mmHg in 13 (15%) patients, a change in MAP, 10

Discussion
Interventions are urgently needed to improve the recognition and treatment of severe sepsis in SSA which suffers the highest proportion of global sepsis and associated mortality [4,6]. In the first study of its kind in SSA, we show that there was a paucity of vital signs monitoring which decreased over the course of hospitalization for adult patients admitted with severe sepsis to a regional referral hospital in Uganda. In resource-rich settings, early identification of deteriorating patients has been shown to improve outcomes through initiation of targeted interventions [1,3,17]. However, early identification requires frequent monitoring and the response to intravenous fluids and antibiotics requires follow-up evaluation [1,13,14]. In our study population, there was    a high mortality rate suggesting that there were opportunities for additional interventions after admission which may have improved outcomes. The reasons for the infrequent monitoring and poor outcomes are complex and likely related to a low clinical staff-topatient ratio as well as a lack of diagnostic and therapeutic resources.
A comprehensive response to clinical deterioration after admission for severe sepsis requires an afferent limb to identify the patient in possible crisis and to initiate an alert, and an efferent limb to rapidly evaluate and treat the patient. In resource-rich settings, this is often accomplished through automated track-andtrigger protocols that prompt notification of a rapid response team when a predetermined threshold of clinical risk is met [17]. However, even when such systems are in place, there is often a failure of the afferent limb of resuscitation [18]. Although we are not aware of any formalized rapid response teams in government referral hospitals in SSA, a previous study in Uganda showed that targeted monitoring and resuscitation of adults with severe sepsis at admission can improve outcomes [7]. We found that patients with lower admission BP received a greater volume of fluid resuscitation compared to those who had a higher admission BP. This finding suggests the existence of afferent and efferent resuscitation limbs which could be optimized to improve outcomes within the context of a resource-limited environment.
The association between increased frequency of vital signs monitoring and mortality in our study implies that there was clinical concern about patients who ultimately succumbed to severe sepsis. The severity of illness at admission, as measured by aggregate organ dysfunction or CRB-65 score, did not predict increased monitoring, but an increased respiratory rate at admission did predict both increased monitoring and mortality. Therefore, clinical concern may have originated from the admission respiratory rate. It is also possible that there was a causal relationship between increased vital signs monitoring and increased in-hospital mortality mediated through unnecessary or potentially harmful interventions [19,20]. For example, the recently published FEAST trial revealed increased mortality from fluid resuscitation in children with severe sepsis in East Africa [21]. It is therefore possible that unmeasured interventions led to greater mortality among our study patients who were more frequently monitored. However, since the overall frequency of monitoring in our study was low, we believe it was sub-optimal rather than deleterious.
Increased vital signs monitoring should ideally strengthen the afferent limb of resuscitation and improve patient outcomes. In our study, the afferent limb of resuscitation was weakened by a median frequency of vital signs measurement of only #1.1 per day. In contrast, if the WHO's IMAI guidelines were followed, some patients would have had 24 BP measurements in the first day of admission [13,14]. Given our findings, such a high frequency of vital signs monitoring may not be feasible in similarly resource constrained settings. In our study, only 14% of patients received 3 BP measurements in the first 24 hours.
Despite the relationship between monitoring frequency and mortality, there was a disconnect between the afferent and efferent limbs of resuscitation. For example, patients with a high respiratory rate had an increased risk of death and were monitored more frequently, but received no more fluids than the rest of the study population. Additionally, patients with lower admission BP or a high CRB-65 score received more fluid than the rest of the study population, but the volume of fluid given was small and they experienced similar mortality as those with higher admission BP. We interpret these findings to mean that clinicians were able to identify patients at risk of death but they were unable to prevent the progression to death through available interventions.
There are several possible reasons for the low frequency of vital signs monitoring in our study. First, intensive monitoring requires both human and material resources that are rarely available in resource-limited settings. Since we retrospectively diagnosed study patients with severe sepsis, clinicians may also not have been aware at the time of admission that patients had severe sepsis and required close monitoring. Even if patients were diagnosed with sepsis, those who presented with higher BPs were less likely to receive intravenous fluids. Clinicians may have thought these patients were clinically stable and did not require close monitoring or fluid resuscitation. Patients with normal BPs may have had other indicators of poor perfusion which were not detected in this resource-limited setting due in part to the lack of laboratory infrastructure [22]. For example, serum lactate is associated with mortality in severe sepsis independent of organ failure and overt shock [23]. A previous study from Uganda showed that point-ofcare whole blood lactate testing predicted mortality in patients with severe sepsis [24]. Important predictors of mortality included CRB-65 score, number of organ failures, admission respiratory rate, anemia, and HIV infection.
This study did have limitations. Several patient charts were not available for review, so some patients admitted with severe sepsis during the investigated time period could have been missed. However, this was not expected to result in bias as the reason for missing charts was unlikely to be related to the frequency of vital signs monitoring or mortality. The low frequency of monitoring limited our ability to interpret the observed association between vital signs monitoring and mortality. Nonetheless, our findings reveal important deficiencies in the monitoring and resuscitation of patients with severe sepsis.

Conclusions
Clinicians in areas with a high burden of infection and limited resources are frequently faced with critically ill patients with severe sepsis and septic shock. A high degree of suspicion for severe sepsis is needed for timely resuscitation of these patients. In this study, patients who eventually succumbed to sepsis raised a higher level of clinical concern at admission as evidenced by higher monitoring rates, but deficiencies in both the afferent and efferent limbs of resuscitation may have led to poor outcomes. A monitoring strategy that employs resource appropriate clinical early warning scores such as CRB-65 may help clinicians in austere environments to identify patients with severe sepsis that require rapid interventions. Accordingly, we recommend further study of predictors of severe sepsis at admission, the association between close monitoring and death, and the optimal monitoring rate for patients with severe sepsis. Although monitoring throughout hospitalization is important, given the precipitous rate of mortality observed in our study, a focus on the immediate post-admission period should be emphasized.