Predicting haemoglobin deferral using machine learning models: Can we use the same prediction model across countries?

BACKGROUND AND OBJECTIVES: Personalized donation strategies based on haemoglobin (Hb) prediction models may reduce Hb deferrals and hence costs of donation, meanwhile improving commitment of donors. We previously found that prediction models perform better in validation data with a high Hb deferral rate. We therefore investigate how Hb deferral prediction models perform when exchanged with other blood establishments. MATERIALS AND METHODS: Donation data from the past 5 years from random samples of 10,000 donors from Australia, Belgium, Finland, the Netherlands and South Africa were used to fit random forest models for Hb deferral prediction. Trained models were exchanged between blood establishments. Model performance was evaluated using the area under the precision-recall curve (AUPR). Variable importance was assessed using SHapley Additive exPlanations (SHAP) values. RESULTS: Across the validation datasets and exchanged models, the AUPR ranged from 0.05 to 0.43. Exchanged models performed similarly within validation datasets, irrespective of the origin of the training data. Apart from subtle differences, the importance of most predictor variables was similar in all trained models. CONCLUSION: Our results suggest that Hb deferral prediction models trained in different blood establishments perform similarly within different validation datasets, regardless of the deferral rate of their training data. Models learn similar associations in different blood establishments.

A case-control study of risk factors for incident hepatitis B virus infection in South African blood donors.

OBJECTIVES: Hepatitis B virus (HBV) infection remains a global health problem. Risk factors for HBV infection are usually assessed in prevalent rather than incident infections. To identify demographic and behavioral risks associated with incident HBV among South African blood donors. METHODS: A case-control study was performed between November 2014 and January 2018. Cases were blood donors testing positive for HBV DNA with or without hepatitis B surface antigen but negative for antibody to hepatitis B core antigen. Participants completed an audio computer-assisted structured interview on exposures during the previous 6 months. Sex-specific multivariable logistic regression yielded independent associations between risks and HBV infection. RESULTS: 56 females and 37 males with incident HBV were compared to 438 female and 439 male controls, respectively. For females, risk factors were accepting money or goods for sex, using agents to prepare one's anus prior to anal sex, penetrating injury, non-Black race, and lower educational status. Men reporting homosexual or bisexual orientation or sex with other men, previous injury, referral for HBV testing, or lack of medical insurance were at increased risk. For both sexes, having more than two male sexual partners increased risk. CONCLUSIONS: Sexual behaviors predominated over parenteral exposures as risks for incident HBV in both female and male blood donors.

The evolution and interpretation of seroprevalence of SARS-CoV-2 antibodies among South African blood donors from the Beta to Omicron variant-driven waves.

BACKGROUND AND OBJECTIVES: Confirmed COVID-19 diagnoses underestimate the total number of infections. Blood donors can provide representative seroprevalence estimates, which can be leveraged into reasonable estimates of total infection counts and infection fatality rate (IFR). MATERIALS AND METHODS: Blood donors who donated after each of three epidemic waves (Beta, Delta and first Omicron waves) were tested for anti-SARS-CoV-2 nucleocapsid antibodies using the Roche Elecsys anti-SARS-CoV-2 total immunoglobulin assay. Roche Elecsys anti-spike antibody testing was done for the post-Omicron sampling. Prevalence of antibodies was estimated by age, sex, race and province and compared to official case reporting. Province and age group-specific IFRs were estimated using external excess mortality estimates. RESULTS: The nationally weighted anti-nucleocapsid seroprevalence estimates after the Beta, Delta and Omicron waves were 47% (46.2%-48.6%), 71% (68.8%-73.5%) and 87% (85.5%-88.4%), respectively. There was no variation by age and sex, but there were statistically and epidemiologically significant differences by province (except at the latest time point) and race. There was a 13-fold higher seroprevalence than confirmed case counts at the first time point. Age-dependent IFR roughly doubled for every 10 years of age increase over 6 decades from 0.014% in children to 6.793% in octogenarians. CONCLUSION: Discrepancies were found between seroprevalence and confirmed case counts. High seroprevalence rates found among Black African donors can be ascribed to historical inequities. Our IFR estimates were useful in refining previous large disagreements about the severity of the epidemic in South Africa. Blood donor-based serosurveys provided a valuable and efficient way to provide near real-time monitoring of the ongoing SARS-CoV-2 outbreak.

Association of ABO and RhD blood groups with the risk of HIV infection.

Naturally occurring antibodies against ABO antigens present in human sera have been shown to neutralize ABO-expressing HIV in vitro. We investigated associations between ABO and RhD blood groups and HIV infection among blood donors from all blood collection centers in eight of South Africa's nine provinces. Whole blood donations collected from first time donors between January 2012 and September 2016 were tested for HIV RNA by nucleic acid testing and HIV antibody using third generation serology assays. ABO and RhD blood types were determined using automated technology. Odds ratios for the association between HIV positivity and ABO and RhD phenotypes were calculated using multivariable logistic regression analysis. We analyzed 515,945 first time blood donors and the overall HIV prevalence was 1.12% (n = 5790). After multivariable adjustment, HIV infection was weakly associated with RhD positive phenotype (OR = 1.15, 95% CI 1.00-1.33) but not with ABO blood group. The observed association with RhD positive phenotype was marginal and likely due to residual confounding by racial group but could serve to generate hypotheses for further studies.

Estimates of prevalence of anti-SARS-CoV-2 antibodies among blood donors in South Africa in March 2022.

In line with previous instalments of analysis from this ongoing study to monitor 'Covid Seroprevalence' among blood donors in South Africa, we report on an analysis of 3395 samples obtained in mid-March 2022 from all provinces of South Africa - a timepoint just after the fourth (primarily omicron) wave of infections. As in our previous analyses, we see no evidence of age and sex dependence of prevalence, but significant variation by race. Differences between provinces have largely disappeared, as prevalence appears to have saturated. In contrast to previous estimates from this study, which reported only prevalence of anti-nucleocapsid antibodies, this present work also reports results from testing for anti-spike antibodies. This addition allows us to categorise those donors whose only antibodies are from vaccination. Our race-weighted national extrapolation is that 98% of South Africans have some antibodies, noting that 10% have anti-spike antibodies but not anti-nucleocapsid antibodies - a reasonable proxy for having both 1) been vaccinated and 2) avoided infection.

Convalescent plasma in the treatment of moderate to severe COVID-19 pneumonia: a randomized controlled trial (PROTECT-Patient Trial).

There is a need for effective therapy for COVID-19 pneumonia. Convalescent plasma has antiviral activity and early observational studies suggested benefit in reducing COVID-19 severity. We investigated the safety and efficacy of convalescent plasma in hospitalized patients with COVID-19 in a population with a high HIV prevalence and where few therapeutic options were available. We performed a double-blinded, multicenter, randomized controlled trial in one private and three public sector hospitals in South Africa. Adult participants with COVID-19 pneumonia requiring non-invasive oxygen were randomized 1:1 to receive a single transfusion of 200 mL of either convalescent plasma or 0.9% saline solution. The primary outcome measure was hospital discharge and/or improvement of ≥ 2 points on the World Health Organisation Blueprint Ordinal Scale for Clinical Improvement by day 28 of enrolment. The trial was stopped early for futility by the Data and Safety Monitoring Board. 103 participants, including 21 HIV positive individuals, were randomized at the time of premature trial termination: 52 in the convalescent plasma and 51 in the placebo group. The primary outcome occurred in 31 participants in the convalescent plasma group and and 32 participants in the placebo group (relative risk 1.03 (95% CI 0.77 to 1.38). Two grade 1 transfusion-related adverse events occurred. Participants who improved clinically received convalescent plasma with a higher median anti-SARS-CoV-2 neutralizing antibody titre compared with those who did not (298 versus 205 AU/mL). Our study contributes additional evidence for recommendations against the use of convalescent plasma for COVID-19 pneumonia. Safety and feasibility in this population supports future investigation for other indications.

Estimates of prevalence of anti-SARS-CoV-2 antibodies among blood donors in eight provinces of South Africa in November 2021.

In line with previous instalments of analysis from this ongoing study to monitor 'Covid Seroprevalence' among blood donors in South Africa, we report on analysis of 3395 samples obtained from 8 to 12 November 2021 in all provinces of South Africa except the Western Cape. As in our previous analyses, we see no evidence of age and sex dependence of prevalence, but substantial variation by province, and by race within each province, from which we generated provincial total point estimates (EC-74%; FS-75%; GP-68%; ZN-73%; LP-66; MP-73%; NC-63%; NW-81% ) and a 'South Africa minus Western Cape' national prevalence estimate of 71% (95%CI 69-74%). We note that sample collection occurred just before the omicron variant driven wave in South Africa, but otherwise present these results without significant interpretation.

Characterising differences in red blood cell usage patterns between healthcare sectors in South Africa: 2014-2019.

BACKGROUND: South Africa aims to transition from a two-tiered healthcare system (public and private) to universal health coverage. Data on red blood cell (RBC) product usage reveal disparities between the sectors. Blood transfusion services further need to understand differing disease profiles and transfusion prescribing practices between the sectors to ensure blood security should the transition to a two-tiered health system come to fruition. MATERIALS AND METHODS: Operational data for public and private healthcare RBC requests between 1 January 2014 and 31 March 2019, obtained from the South African National Blood Service (SANBS), were retrospectively analysed. Sector-specific demographic and utilisation trends were compared for the dominant clinical disciplines. Pre-transfusion haemoglobin (Hb) patterns were also delineated for 2018. RESULTS: Between 2014 and 2019, 2,356,411 public and private sector RBC transfusion events resulted in the issue of 4,020,094 RBC units (1,553,159 transfusion events and 2,495,054 units within the public sector versus 803,282 transfusion events and 1,525,040 units in private). The dominant clinical disciplines within the public sector were Medical (32.9%), Gynaecology/Obstetrics (27.3%), General Surgery (13.6%), and Paediatrics (including Paediatric Surgery) (6.5%), compared to Intensive Care Units (33.2%), Medical (28.3%), General Surgery (10.4%), and Haematology/Oncology (8.3%) in the private sector. Median pre-transfusion Hb values for 2018 were lower in the public than in the private sector: 6.9 g/dL public sector versus 8 g/dL private sector. DISCUSSION: Clinical drivers of RBC usage within the public and private healthcare sectors in South Africa differ significantly. Disparate pre-transfusion Hb between the sectors are likely due to differing disease profiles and severity, as well as differences in practice in prescribing transfusions. Implementation of a nationally co-ordinated Patient Blood Management programme may help to address these disparities and help ensure a sustainable blood transfusion system.

Re-engineering the medical assessment of blood donors in South Africa: The balance between supply and safety.

BACKGROUND: The balance between ensuring blood donor and recipient safety while maintaining a sufficient blood supply can be affected by excessive deferral of blood donors. In 2018, a biannual regulatory review of donor eligibility criteria provided the South African National Blood Service (SANBS) the opportunity to review the existing criteria. Changes to these criteria were implemented in April 2019 after an extensive review. STUDY DESIGN AND METHODS: We conducted a cross-sectional study of SANBS whole-blood donor presentations to determine the impact of the changed donor eligibility criteria on deferrals and blood safety. We compared donor presentations, deferrals, and HIV-positive cases for the 12-month period (April 2019-March 2020) after the implementation of the updated donor eligibility criteria to those of the previous year. RESULTS: Of the 2,112,917 donor presentations, 51.1% (1079506) occurred in the post-implementation study period. Overall, deferrals decreased from 18.6% to 14.5%, whereas HIV-positive donations increased by 0.03%. A multivariable logistic regression analysis adjusted for sex, age, geographical location, donor, and clinic type showed significantly lower odds of deferral (OR 0.70; 95% CI: 0.69-0.70) and greater odds of HIV-positive cases in the study period than those in the control period (OR 1.17; 95% CI: 1.10-1.25). CONCLUSION: We confirmed that the change in donor eligibility criteria was associated with a decrease in deferrals and an increase in the country's blood supply. The impact of the increased number of HIV-positive donations on blood safety in a country performing individual donation nucleic acid amplification testing requires further investigation.

Rapid and Successful Implementation of a COVID-19 Convalescent Plasma Programme-The South African Experience.

BACKGROUND: COVID-19 convalescent plasma (CCP) has been considered internationally as a treatment option for COVID-19. CCP refers to plasma collected from donors who have recovered from and made antibodies to SARS-CoV-2. To date, convalescent plasma has not been collected in South Africa. As other investigational therapies and vaccination were not widely accessible, there was an urgent need to implement a CCP manufacture programme to service South Africans. METHODS: The South African National Blood Service and the Western Cape Blood Service implemented a CCP programme that included CCP collection, processing, testing and storage. CCP units were tested for SARS-CoV-2 Spike ELISA and neutralising antibodies and routine blood transfusion parameters. CCP units from previously pregnant females were tested for anti-HLA and anti-HNA antibodies. RESULTS: A total of 987 CCP units were collected from 243 donors, with a median of three donations per donor. Half of the CCP units had neutralising antibody titres of >1:160. One CCP unit was positive on the TPHA serology. All CCP units tested for anti-HLA antibodies were positive. CONCLUSION: Within three months of the first COVID-19 diagnosis in South Africa, a fully operational CCP programme was set up across South Africa. The infrastructure and skills implemented will likely benefit South Africans in this and future pandemics.

An assessment of hepatitis B virus prevalence in South African young blood donors born after the implementation of the infant hepatitis B virus immunization program: Implications for transfusion safety.

BACKGROUND: The prevalence of hepatitis B surface antigen is estimated to be 6.7% in the South African population and in April 1995 the nation introduced universal hepatitis B virus (HBV) vaccination for newborns and infants. We studied the temporal association of this program with HBV prevalence in young blood donors and the contemporary HBV incidence and residual risk of transfusion-transmitted HBV infection (TT-HBV). METHODS: We used blood donation data from January 2011 to December 2019. Estimation of HBV prevalence donations made by first-time blood donors were analyzed by birth cohort and covariates. To estimate the incidence and residual risk of TT-HBV, mathematical models used data from both first time and repeat donors. RESULTS: HBV prevalence in first-time donors decreased from 0.84% (95% confidence interval [CI] 0.78-0.90) in 2011 to 0.66% (95% CI 0.61-0.70) in 2019. The post-1995 birth cohort had a significantly lower HBV prevalence of 0.14% (95% CI 0.13-0.15) than the pre-1985 birth cohort of 1.29% (95% CI 1.25-1.33) and the odds of HBV infection were reduced in a multivariable model (odds ratio [OR] = 0.28, 95% CI 0.24-0.34). The residual risk of TT-HBV occurring from window-period, occult, and possible vaccine breakthrough infections were estimated at 36.9, 5.8, and 2.2 per million red blood cell transfusions, respectively. CONCLUSION: Donors born after the start of routine HBV immunization had significantly lower prevalence of HBV infection, supporting the effectiveness of the vaccination program. The contemporary residual risk of TT-HBV has decreased and should decline further as more vaccinated young people join the donor pool.

Prevalence of anti-SARS-CoV-2 antibodies among blood donors in Northern Cape, KwaZulu-Natal, Eastern Cape, and Free State provinces of South Africa in January 2021.

Background: Population-level estimates of prevalence of anti-SARS-CoV-2 antibody positivity (seroprevalence) is a crucial epidemiological indicator for tracking the Covid-19 epidemic. Such data are in short supply, both internationally and in South Africa. The South African blood services (the South African National Blood Service, SANBS and the Western Cape Blood Service, WCBS) are coordinating a nationally representative survey of blood donors, which it is hoped can become a cost-effective surveillance method with validity for community-level seroprevalence estimation. Methods: Leveraging existing arrangements, SANBS human research ethics committee permission was obtained to test blood donations collected on predefined days (7th, 10th, 12th, 15th, 20th, 23th and 25th January) for anti-SARS-CoV-2 antibodies, using the Roche Elecsys Anti-SARS-CoV-2 assay on the cobas e411 platform currently available in the blood services' donation testing laboratories. Using standard methods, prevalence analysis was done by province, age and race, allowing age to be regarded as either a continuous or categorical variable. Testing was performed in the Eastern Cape (EC), Free State (FS), KwaZulu Natal (ZN) and Northern Cape (NC) provinces. Results: We report on data from 4858 donors - 1457 in EC; 463 in NC; 831 in FS and 2107 in ZN. Prevalence varied substantially across race groups and between provinces, with seroprevalence among Black donors consistently several times higher than among White donors, and the other main population groups (Coloured and Asian) not consistently represented in all provinces. There is no clear evidence that seroprevalence among donors varies by age. Weighted net estimates of prevalence (in the core age range 15-69) by province (compared with official clinically-confirmed COVID-19 case rates in mid-January 2021) are: EC-63%(2.8%), NC-32%(2.2%), FS-46%(2.4%), and ZN-52%(2.4%). Conclusions: Our study demonstrates substantial differences in dissemination of SARS-CoV-2 infection between different race groups, most likely explained by historically based differences in socio-economic status and housing conditions. As has been seen in other areas, even such high seroprevalence does not guarantee population-level immunity against new outbreaks - probably due to viral evolution and waning of antibody neutralization. Despite its limitations, notably a 'healthy donor' effect, it seems plausible that these estimates are reasonably generalisable to actual population level anti-SARS-CoV-2 seroprevalence, but should be further verified.

HIV incidence in South African blood donors from 2012 to 2016: a comparison of estimation methods.

BACKGROUND: Measuring incidence is important for monitoring and maintaining the safety of the blood supply. Blood collected from repeat-donors has provided the opportunity to follow blood donors over time and has been used to estimate the incidence of viral infections. These incidence estimates have been extrapolated to first-time donors using the ratio of NAT yield cases in first-time versus repeat-donors. We describe a model to estimate incidence in first-time donors using the limiting antigen (LAg) avidity assay and compare its results with those from established models. METHODS: HIV-positive first-time donations were tested for recency using the LAg assay. Three models were compared; incidence estimated for (1) first-time donors using LAg avidity, (2) first-time and repeat-donors separately using the NAT yield window period (WP) model and (3) repeat-donors using the incidence/WP model. RESULTS: HIV incidence in first-time donors was estimated at 3·32 (CI 3·11, 3·55) and 3·81 (CI 3·07, 4·73) per 1000 PY using the LAg assay and NAT yield WP models, respectively. Incidence in repeat-donors was between 2·0- and 2·5-fold lower than in first-time donors estimated at 1·56 (CI 1·37, 1·77) and 1·94 (CI 1·86-2·01) per 1000 PY using the NAT yield/WP and incidence/WP models, respectively. CONCLUSION: Testing HIV-positive donations using the LAg assay provides a reliable method to estimate incidence in first-time donors for countries that collect the majority of blood from first-time donors and do not screen with NAT.

Using a motivator and deterrent questionnaire to predict actual donation return behavior among first-time African-origin blood donors.

BACKGROUND: In South Africa, success in recruiting first-time African-origin blood donors at low risk for HIV infection has been tempered by lower-than-anticipated return rates. We studied whether self-reported motivators and deterrents could predict actual donor return among first-time donors. METHODS: We administered a questionnaire on blood donation motivators and deterrents within 2 months after donation to a cohort of first-time African-origin blood donors in 2014 and followed their return behavior for 1 year. Survey responses associated with return within 1 year were assessed using separate multivariate logistic regression models for motivators and deterrents. RESULTS: A total of 2902 first-time African origin donors with median age of 23 years and female predominance (59%) were included. Within 1 year 1574 (54% of donors) attempted at least one additional donation. Donors were more likely to return if they strongly agreed with the statement "Blood donation is an easy way to make a difference" (odds ratio [OR], 2.0; 95% confidence interval [CI], 1.3-2.9) and "I donated in response to advertisements on the radio, TV or newspaper" (OR, 1.6; 95% CI, 1.2-2.1). English language preference and nonstudent employment status were also associated with donor return. Poor customer service (OR, 0.45; 95% CI, 0.28-0.71), geographic region, and nonstudent employment status were associated with donor return in the deterrent model. CONCLUSIONS: Prosocial motivators and marketing communications predicted return, while poor customer experience predicted nonreturn. This novel design allowed us to study the link between donation motivators and deterrents and actual return for donation in the African context.

Discovery of False Elite Controllers: HIV Antibody-Positive RNA-Negative Blood Donors Found To Be on Antiretroviral Therapy.

BACKGROUND: An increase in potential HIV elite controllers (EC) and anecdotal reports of antiretroviral therapy (ART) use among South African blood donors led us to verify EC status. METHODS: Stored plasma samples from potential EC were tested for ART drugs. Demographic and temporal associations were examined using multivariable logistic regression. RESULTS: Of 226 potential EC, 150 (66.4%) had detectable ART with increasing prevalence by year (OR = 7.57 for 2016 vs 2010, 95% confidence interval, 1.96-32.17). DISCUSSION: False presumptive EC status due to undisclosed ART represents a growing proportion of potential EC donors in South Africa coincident with the country's ART rollout.

Assessment of HIV transfusion transmission risk in South Africa: a 10-year analysis following implementation of individual donation nucleic acid amplification technology testing and donor demographics eligibility changes.

BACKGROUND: In 1998 we estimated that 34/million infectious window period donations were entering the blood supply at the South African National Blood Service. Selective use of donations based on donor race-ethnicity reduced this risk to 26/million donations but was deemed unethical. Consequently, in 2005 South African National Blood Service eliminated race-ethnicity-based collection policies and implemented individual-donation nucleic acid testing (ID-NAT). We describe the change in donor base demographics, human immunodeficiency virus (HIV) detection rates, and transfusion-transmissible HIV risk. STUDY DESIGN AND METHODS: In ten years 7.7 million donations were tested for anti-HIV and HIV RNA. Number of donations, HIV prevalence, ID-NAT yield rate, serology yield rate and residual transfusion-transmissible HIV risk were analyzed by donor type, race-ethnicity, age, and sex. Multiple regression analysis was performed to investigate the determinants of HIV-positive and nucleic acid testing yield donations. RESULTS: The combined strategy of increasing donations from black donors and implementing ID-NAT increased the proportion of donations from black donors from 6% in 2005 to 30% in 2015 (p < 0.00001), and reduced the transfusion-transmissible risk from 24 to 13 per million transfusions. ID-NAT interdicted 481 (1:16,100) seronegative window period donations, while one transfusion-transmissible case (0.13 per million) was documented. Race-ethnicity and donor type were highly significant predictors of HIV positivity, with adjusted odds ratio for first-time donors of 12.5 (95% confidence interval, 11.9-13.1) and for black race-ethnicity of 31.1 (95% confidence interval, 28.9-33.4). The proportion of serology yields among HIV-infected donors increased from 0.27% to 2.4%. CONCLUSION: ID-NAT enabled the South African National Blood Service to increase the number of donations from black donors fivefold while enhancing the safety of the blood supply.

The iron status of South African blood donors: balancing donor safety and blood demand.

BACKGROUND: Several studies in developed countries have demonstrated high levels of iron deficiency (ID) among blood donors. There is a paucity of data for developing countries where blood shortages remain a major concern. STUDY DESIGN AND METHODS: A total of 4412 donors were enrolled in the study. Specimens were collected for full blood count, iron, transferrin saturation, and ferritin assessment. Donor demographics were recorded. ID was indicated by a ferritin level of less than 20 ng/mL for men and less than 12 ng/mL for women. Anemia was defined as hemoglobin levels less than 12.5 g/dL. Regression models for predictors of ID were developed. RESULTS: A total of 17.5% of all donors had ID, with 16.3% prevalence in women and 18.6% in men. Low hemoglobin had the highest association with ID (adjusted odds ratio [AOR], 11.078; 95% confidence interval [CI], 7.915-15.505); male donors had twice the odds of ID compared to female donors (AOR, 2.501; 95% CI, 1.964-3.185), while increasing age was associated with lower odds (AOD, 0.965; 95% CI, 0.956-0.975). Among male donors, an interdonation interval of less than 3 months (AOR, 2.679; 95% CI, 1.929-3.720) was associated with ID. Compared to other females combined, colored female donors (AOR, 2.335; 95% CI, 1.310-4.160) had higher odds and black female donors (AOR, 0.559; 95% CI, 0.369-0.845) lower odds of ID. CONCLUSION: ID is common among South African donors; low hemoglobin, gender, ethnicity, and past donation history is independently associated with ID. Recommendations aimed at protecting donor health may increase blood shortages in South Africa.

Use of Blood Donor Screening to Monitor Prevalence of HIV and Hepatitis B and C Viruses, South Africa.

Among 397,640 first-time blood donors screened in South Africa during 2012-2015, HIV prevalence was 1.13%, hepatitis B virus prevalence 0.66%, and hepatitis C virus prevalence 0.03%. Findings of note were a high HIV prevalence in Mpumalanga Province and the near absence of hepatitis C virus nationwide.