Severe Sepsis in Uganda— Findings and Future Directions from the PRISM-U Study Group W. Michael Scheld, MD University of Virginia May 2008 PRISM-U: Promoting ResourceLimited Interventions for Sepsis Management in Uganda Co-Investigators: Shevin T Jacob (UVA) Patrick Banura (Masaka Hospital) Christopher C Moore (UVA) David Meya (IDI) Steven J Reynolds (RHSP/NIH) Pius Opendi (RHSP) Relana Pinkerton (UVA) Principal Investigators: Nathan Kenya-Mugisha (MOH) Harriet Mayanja-Kizza (Makerere) W Michael Scheld (UVA) Outline Rationale for PRISM-U1 Data from PRISM-U1 Future directions with PRISM-U2 Genre: Progressive death metal Lyrical themes: Human nature, life and death Discography: The Town of Death, Subjugate the Chaos RP Wenzel. NEJM 2002; 347 (13): 966-967 Severe sepsis—an important cause of HIV-associated mortality in Africa? Increased bacteremia in HIV-infected patients 3 cross sectional studies in Kenya showed that the most common organisms isolated were: • non-typhoidal Salmonella • Streptococcus pneumoniae • Mycobacterium tuberculosis High mortality associated with bacteremia in HIV-infected patients. Prospective observational study of febrile adult patients admitted to Mulago Hospital (n=299) • 24% had bacteremia • 76% were HIV-infected Inpatient mortality among bacteremic pts was 28% Mortality in bacteremic patients likely resulted from severe sepsis/ septic shock Sepsis and septic shock were amongst the three most common causes of mortality in a Ethiopian cohort of HIVinfected patients. Arthur G, et al. Clin Infect Dis 2001; 33(2): 248-56. FN Ssali, et al. Journal of Acquired Immune Deficiency Syndromes & Human Retrovirology 1998; 19(5): 484-489. A Bane, et al. Ethiop Med J. 2003; 41(2): 131-40. Management of sepsis in Africa No prospective studies on the management and outcome of severely septic patients in SSA Zambian study showed suboptimal management of sepsis associated with high mortality. Retrospective chart review of hospitalized Zambian patients with fever and hypotension (n=302) 86% (79/92) of hypotensive septic patients received no intravenous fluid resuscitation 35% in-hospital mortality Tanzanian study showed increased mortality seen when empiric antibiotics discordant with sensitivity profiles. Is there a potential for HIV-associated mortality reduction through improved sepsis management? C.Theodosis, et al. Abstract presented at International AIDS Conference, Toronto, 2006. B Blomberg, et al. BMC Infect Dis 2007; 7:43. Rounds at Mulago Hospital Patient Lying on the Floor PRISM-U, Part 1 Prospective observational study investigating the management and outcomes of patients admitted to medical wards with severe sepsis Study details Time frame: July 2006 to November 2006 2 sites Mulago Hospital Masaka Regional Referral Hospital Inclusion Suspected infection 2 of the following: • Body temperature >37.5°C or <35.5°C • Heart rate >90 beats/min • Respiratory rate >20 breaths/min Systolic Blood Pressure ≤ 100 mmHg Exclusion Age <18yo Acute cerebrovascular event GI bleed Need for triage to surgery or OB/GYN ward Methods: Ethical approval University of Virginia Institutional Review Board Makerere University Faculty of Medicine Research Ethics Committee Mulago Hospital Office of Director Infectious Disease Institute Scientific Research Committee Uganda National Council of Science and Technology Results: Patient characteristics N: 385 eligible, 382 enrolled 250 patients from Mulago Hospital 132 patients from Masaka Hospital Sex: 59.2% female Age: mean, 34 yo (range 18 to 80) HIV: Seropositive: 84.9% (320/377) CD4: median, 52 (range 1-999) On HAART: 11.9% (38/320) Unaware of serostatus: 32.5% (104/320) Mean KPS, admission: 46 (range 10-90) Fulfilled ≥ 3 SIRS criteria: 83.2% Most frequent chief complaints: Fever: 39.2% Cough: 19.6% Diarrhea: 11.0% Results: Hospital comparisons Mulago Masaka P value Age, mean 34.34 ± .67 35.56 ± 1.07 .310 Female 59.6% 58.3% .811 HIV prevalence 85.9% (n=248) 82.9% (n=132) 0.449 CD4, mean 82.24 ± 7.56 (n=212) 146.71 ± 17.95 (n=107) <0.001 Aware of Status 72.3% (n=212) 57.9% (n=107) 0.013 On HAART 13.1% (n=213) 9.3% (n=107) .322 Admission KPS 43.24 ± 1.021 52.29 ± 1.399 <0.001 Fulfills ≥ 3 SIRS criteria* 84.8% (n=250) 80.2% (n=131) .249 Portable lactate (mmol/L)* 4.11 ± .31 3.70 ± .22 0.75 + malaria blood smear 9.6% (n=250) 20.9% (n=129) .002 Results: Microbiology Bacteremia was found in 18.9% of 381 patients. Most common organisms were: Mycobacterium tuberculosis (22.1%)* non-typhoidal Salmonella (17.1%) Staphylococcus aureus (8.5%) Streptococcus pneumoniae (5.1%) Cryptococcus neoformans (4.3%) Similar findings in 1997 Mulago study by Ssali et al. Results: Mycobacterial cultures 22.1% (n=55/249) had mycobacteremia 45.5% (n=25 of 55) had MTB 54.5% (n=30 of 55) had NTM Are the NTM isolates a true pathogen or laboratory contamination? Studies from Uganda and other SSA countries report MAI to be a rare pathogen in AIDS patients; other NTM rarely reported Inpatient mortality increased in patients with MTB bacteremia compared to patients with negative mycobacterial cultures (69.6% vs. 44.4%, P=0.024) No significant differences in mortality between patients with NTM and negative cultures (48% vs. 44.4%, P=0.755) CF Gilks, et al. J Acquir Immune Defic Syndr Hum Retrovirol 1995;8(2):195-8. DO Okello, et al. J Infect Dis 1990;162(1):208-10. Combined mortality for patients with negative mycobacterial, MTB and NTM cultures Results: Outcomes Median length of hospitalization (days): Lost to follow-up (after discharge): 6 (range, 1-55) 14.8% (43/290) Mortality In-hospital: 23.7% (n=380) Post discharge, 30-day mortality: 22.3% (n=248) Combined (in-hospital + 30-day mortality): 43.0% (n=337) Results: Clinical predictors Logistic regression used to assess the predictive value of SIRS clinical variables, morbidity assessment scales (eg, KPS and GCS), and laboratory results for in-hospital and post discharge mortality. In-hospital mortality model: Significant independent predictors: • Temperature at 6 hrs post admission (P=0.04) • Admission KPS (P=0.003) Post discharge mortality model: Significant independent predictors: • Discharge KPS (P=0.003) • CD4 count (P=0.006) Karnofsky Performance Scale A subset of 72 patients from the Mulago Hospital site were assessed to determine the utility of portable lactate testing as a predictor of mortality. No patients died if their lactate level was ≤ 2.5 mmol/L. PWBL was positively associated with in-hospital but not outpatient mortality (P<0.001). In-hospital mortality was increased in patients with a PWBL concentration ≥ 4.0 mmol/L vs. < 4.0 mmol/L: 50% vs 7.5%, OR 12.3 [3.5–48.9]; P<0.001 Using multiple logistic regression, standard laboratory serum lactate results were inconsistent and less predictive of mortality than were those of PWBL. Accutrend® portable lactate machine Allows point of care testing Requires minimal technical expertise Requires minimal sample preparation or preservation In-hospital mortality at different PWBL concentrations n = 5/5 n = 10/25 n = 3/18 n = 0/22 Results: Intravenous Fluid Resuscitation Recommended initial administration of fluid: 53% received at least some IV crystalloid within the first hour of presentation 28% waited between 1 to 6 hours for fluid 14% of patients received no fluid within the first 6 hours of presentation. Within first 6 hours of presentation: 20-40 cc/kg as a fluid challenge Subsequent fluid challenges every 30 min to 1hour for refractory hypotension within the first 6 hours Mean volume of fluid: 671 cc % receiving > 1L of fluid: 34.5% % receiving > 2L of fluid: 3.7% Within first 24 hours of presentation: Mean volume of fluid: 1075 cc % receiving > 1L of fluid: 62.7% % receiving > 2L of fluid: 13.7% SurvivingSepsisCampaign. http://www.survivingsepsis.org/bundles/individual_changes/treat_hypotension. Intravenous Fluid Resuscitation: Hospital Comparison Fluid parameter Mulago Hospital Masaka Hospital P-value % receiving ≥ 1L in 1st 6 hrs 42.9% 16.8% <0.001 % receiving ≥ 2L in 1st 6 hrs 5.2% 0.8% 0.03 % receiving ≥ 65.1% 1L in 1st 24 hrs 58.6% 0.21 % receiving ≥ 15.6% 2L in 1st 24 hrs 10.2% 0.14 Intravenous Fluid Resuscitation, cont No overall survival benefit was observed across strata of fluid volume. In patients with any bacteremia (n=71): Trend towards decreased total mortality in those receiving ≥ 1L of fluid vs. < 1L of fluid in the first 6 hours • 36.4% vs. 55.1%, P=0.14 In patients with aerobic bacteremia (n=33): Decreased mortality at 30 days if given ≥ 1L of fluid vs. < 1L of fluid in the first 6 hours • 0% vs. 27.3%, P=0.056 In patients with Gram negative bacteremia (n=25): Decreased total mortality in those receiving ≥ 1L of fluid vs. < 1L of fluid in the first 6 hours • 20% vs 66.7%, P=0.022 PRISM-U1: Summary points Admission for suspected severe sepsis in Uganda is associated with high mortality and late stage HIV infection Microbiology findings similar to other settings in SSA Bacteremia-associated in-hospital mortality in Mulago Hospital has not improved in the last 10 years (28%33%) Chloramphenicol resistance in NTS suggests that it should be avoided for empiric Gram negative therapy Important predictors of death after discharge include KPS at discharge and CD4 count Management of sepsis in this setting is suboptimal with respect to fluid resuscitation and antibiotic administration Future studies need to be done to evaluate approaches to managing critically ill patients in resource constrained settings like Uganda The next step… Major constraints to providing optimal sepsis care: Lack of skilled health workers—on a 50-bed ward where the number of patients can approach 100, the nurse to patient ratio was less than 1:20 Insufficient fluid supplies—the volume of intravenous fluids provided to the ward/day was limited to 20 L There are no current measures in place to decrease the high mortality (~43%) associated with sepsis In Uganda, family members accompany patients for the duration of their hospital stay and are expected to provide limited assistance for daily patient needs such as feeding, bathing and toileting. 91.6% patients in PRISM-U1 accompanied by an attendant Attendants have been used in other settings (eg. Hospice Africa) to provide care usually attributed to nurses Patient attendants may be an untapped resource for contributing to patient care in busy Ugandan hospitals. PRISM-U2: Aims To determine if early aggressive fluid resuscitation improves outcomes among patients with severe sepsis in Uganda (Jacob, Banura) To determine if patient attendants (Musawos) can assist in early fluid resuscitation. To obtain samples for later testing on pathogenesis/pathophysiology. PRISM-U2: Study Samples Serum cytokines (Moore) Blood 16S rDNA PCR (Fredricks) Mycobacterial speciation (Manabe) Buccal swabs; SNPs, WGA (Rich, Concannon) HIV clade and tropism (Yuan) Point of care microfluidic diagnostics (Kane) Discharge Discharge Accident & Emergency Triage Medicine casualty consent Fulfills Inclusion Criteria Ward 3B Ward 4A/ 4B/ 4C Discharge Portable whole blood lactate Death ≤ 2.5 mmol/ L •Admit to surgery ward •Admit to OB/GYN ward Exclude from study Standard arm >2.5 mmol/ L OR Karnofsky ≤ 40 Intervention arm Provide fluids and training to mujanjabi Karnofsky scale at discharge and 30 days THERAPY: •Time from arrival to treatment •Treatment within first 6h? Within first 24? •Amt of fluid within first 6h? First 24? Overall? •Which antibacterial/anti-TB/ anti malarial? Monitor for pulm edema q 1 hr for first 6 hrs INITIAL EVAL: <18 yo, GI bleed, acute cerebrovascular event •Chief complaint, other symptoms, duration of illness •Demographics (age, sex, address, phone #, referral mode) •HIV info: knowledge of serostatus, previous OI, previous HAART, previous OI Rx, WHO clinical stage •Karnofsky scale on admission Collect buccal washings only LABS/STUDIES: Mulago Hospital •WBC, Hb, plts, WBC differential •Albumin •HIV test, CD4 •Malaria smear •Aerobic and Mycobacterial blood cultures •Buccal washings for genomic DNA •Tissue banking: •Cytokine profiles •HIV chemoreceptor and tropism •Molec. analysis for rare organisms OUTCOMES: •In hospital mortality •Length of hospitalization •Out of hospital 30-d survival Thank you! Thank you. PRISM-U2 Methodology: Enrollment criteria Inclusion Patients triaged to the Medical Casualty Units of Mulago Hospital and Masaka Regional Referral Hospital with: • Suspected underlying infection • 2 of the following: • • • Tachycardia > 90 beats per minute Tachypnea > 20 bpm Thermodysregulation (body temperature < 35.5° or > 37.5° C) • Systolic Blood Pressure ≤ 100 mm Hg • Accompanied by an attendant Exclusion Patients presenting to the Accident and Emergency unit of Mulago Hospital or Masaka Regional Referral Hospital with: • < 18 years of age • History of hypervolemic disease state (i.e., congestive heart failure, hepatic failure, nephrotic syndrome) • Signs of fluid overload on physical exam • Acute cerebrovascular event • Gastrointestinal bleed • Need for triage to the Surgical Casualty Unit or Obstetrics and Gynecology Unit PRISM-U2 Methodology: Study details In PRISM-U1, both portable lactate (PL) concentration and admission Karnofsky scale were independent predictors of mortality After initial screening, patients will be further selected based on PL testing using a cut-off value of > 2.5 mmol/L or a Karnofsky score of ≤ 40. Applying these criteria to our preliminary cohort: • 95% (39/41) of excluded patients survived • 50% (28/56) of included patients died 660 severely septic patients will be consecutively enrolled at Mulago and Masaka Hospitals over 7 months 422 patients will be entered in the intervention arm if they have either a PL concentration > 2.5 or a Karnofsky Performance Scale ≤ 40. The remaining 238 patients will have lactate levels ≤ 2.5 and thus will provide only buccal washing for DNA analysis in hypothesis 3. PRISM-U2 Methodology: Intervention arm Attendants will be trained by a study team member to identify when IV fluid bottles are empty. Once resuscitation begins, attendant will alert the medical ward study team medical officer (STMO) that the fluid bottle needs changing. Bottles of normal saline will be provided for appropriate early fluid resuscitation within the first 6 hours of presentation. All patients will receive enough fluid sufficient for a fluid challenge of 30 cc/kg in the first 60 minutes and subsequent fluid challenges for every 60-minute period thereafter. Role of STMO: Available on the ward to assist attendants who need help with their patient. To assess whether fluid bottles are empty every 30 minutes. If bottle empty and STMO not alerted by attendant, STMO will replace fluid bottle and document reason why attendant did not alert him/her. To monitor vital signs every hour (including RR, O2 sats, BP) for the first 6 hours of hospital admission PRISM U2: Risks Fluid overload leading to pulmonary edema. A thorough baseline cardiopulmonary exam will be performed by the monitoring study team member All subjects in the intervention arm will have hourly O2 saturation monitoring and repeat cardiopulmonary exams If pulmonary edema occurs, chest XRAY, nasal cannula oxygen and furosemide will be available for administration if co-investigator assesses that they are clinically warranted. Data Safety Monitoring Board (DSMB) in place to ensure that in-hospital mortality is not higher than expected Venipuncture Risks from Tissue Banking and Genetic Testing All samples will be linked and coded in order to protect from disclosure. PRISM U2: Benefits Laboratory and microbiology tests which otherwise might not be afforded will be available to patients and their treating physician as soon as results are available. Because of the 30-day follow-up contact, ill patients will be asked to return to the hospital for evaluation. Receiving adequate fluid resuscitation has decreased sepsis mortality in other settings— May help to decrease the high mortality in this setting Results from the study may lead to policy changes which increase funding for fluids and health workers—May provide a solution to the current health care worker crisis contributing to high sepsis-associated mortality in this setting. PRISM-U2, Capacity Building Providing health workers with algorithms for the care of critically ill patients Including a PGY student in the study to assist with thesis development and train in laboratory research Teaching students and interns about sepsis management Donating study supplies to respective hospitals after enrollment is complete Acknowledgments PRISM-U Study Team Mulago Hospital • • • • Dr. Angelo Nganizi Dr. Cassim Kalisa Dr. Kasozi Kimuli Mr. Samson Omongot Masaka Hospital • Dr. Francis Ssali • Mr. Patrick Ddikusoka • Sister Scholastica Sekayiba Dr. Jackie Mabweijano (head of casualty, Mulago) Nurses and Red Cross workers of the Mulago and Masaka Casualty Departments Ebenezer Labs (Kampala), AID Child (Masaka), and Uganda Cares (Masaka) for laboratory support Infectious Diseases Institute UVA Center for Global Health—Pfizer Initiative for International Health Dr. Michael Scheld Methods: Data collection Fulfillment of inclusion criteria assessed at A&E Temperature, heart rate, respiratory rate, and blood pressure obtained at enrollment and 6 and 24 hours afterwards. Followed patients from emergency ward to medical wards at both sites until discharge or death. The amount of SIRS criteria fulfilled, IV fluid resuscitation provided and type of empiric antimicrobial agents administered in the first 24 hours were recorded. Laboratory and culture results provided to primary medical team to assist in care of patients. Outcomes measured: Length of hospitalization In-hospital mortality 30-day mortality Methods: Laboratory evaluation Rapid HIV-1 antibody testing and malaria thick smears were performed on site at each hospital. Portable whole blood lactate (PWBL) was performed on whole blood collected by finger stick. Mulago: Complete blood count (CBC) with differentials CD4 counts Serum electrolytes Cortisol Albumin Masaka: CBC with differentials CD4 counts Methods: Microbiology Mulago: Aerobic blood cultures performed at the Makerere University microbiology lab Mycobacterial cultures performed at JCRC • IS6110 PCR amplification used to further identify positive AFB smears as MTB complex or NTM Masaka: Aerobic blood cultures were performed at the nearby Rakai Health Sciences Program laboratory Mycobacterial cultures not performed at this site. Results: Empiric antibiotic administration Fifty-two different empiric antibiotic combinations were used by the admitting doctors caring for the study patients. Antibiotic regimens included ampicillin, penicillin, chloramphenicol, ceftriaxone, gentamicin, metronidazole, TMP-SMX, amoxicillin, ampicillincloxacillin, cefuroxime, and erythromycin. Availability at any given time was random and inconsistent. 85.9% of patients received some empiric antibiotics regimen No mortality benefit seen in patients receiving any random empiric regimen compared to patients receiving no empiric antibiotics • 23.8% vs 22.6%, P=0.849 Empiric antibiotic administration, cont Appropriate antibiotic usage was defined as using an empiric antibiotic regimen to which the isolate was found to be sensitive. Among aerobic cultures where susceptibilities were tested (n=30), Decreased total mortality in patients receiving appropriate vs. inappropriate empiric antibiotics • 18.2% vs 50%, P=0.074 Increased survival for discharged patients who received appropriate therapy vs. inappropriate therapy • 100% vs. 69.2%, P=0.066 95% (19/20) of Salmonella isolates were resistant to chloramphenicol.