The Need for Ethnoracial Equity in Artificial Intelligence for Diabetes Management: Review and Recommendations

There is clear evidence to suggest that diabetes does not affect all populations equally. Among adults living with diabetes, those from ethnoracial minority communities—foreign-born, immigrant, refugee, and culturally marginalized—are at increased risk of poor health outcomes. Artificial intelligence (AI) is actively being researched as a means of improving diabetes management and care; however, several factors may predispose AI to ethnoracial bias. To better understand whether diabetes AI interventions are being designed in an ethnoracially equitable manner, we conducted a secondary analysis of 141 articles included in a 2018 review by Contreras and Vehi entitled “Artificial Intelligence for Diabetes Management and Decision Support: Literature Review.” Two members of our research team independently reviewed each article and selected those reporting ethnoracial data for further analysis. Only 10 articles (7.1%) were ultimately selected for secondary analysis in our case study. Of the 131 excluded articles, 118 (90.1%) failed to mention participants’ ethnic or racial backgrounds. The included articles reported ethnoracial data under various categories, including race (n=6), ethnicity (n=2), race/ethnicity (n=3), and percentage of Caucasian participants (n=1). Among articles specifically reporting race, the average distribution was 69.5% White, 17.1% Black, and 3.7% Asian. Only 2 articles reported inclusion of Native American participants. Given the clear ethnic and racial differences in diabetes biomarkers, prevalence, and outcomes, the inclusion of ethnoracial training data is likely to improve the accuracy of predictive models. Such considerations are imperative in AI-based tools, which are predisposed to negative biases due to their black-box nature and proneness to distributional shift. Based on our findings, we propose a short questionnaire to assess ethnoracial equity in research describing AI-based diabetes interventions. At this unprecedented time in history, AI can either mitigate or exacerbate disparities in health care. Future accounts of the infancy of diabetes AI must reflect our early and decisive action to confront ethnoracial inequities before they are coded into our systems and perpetuate the very biases we aim to eliminate. If we take deliberate and meaningful steps now toward training our algorithms to be ethnoracially inclusive, we can architect innovations in diabetes care that are bound by the diverse fabric of our society.

Respiratory, birth and health economic measures for use with Indigenous Australian infants in a research trial: a modified Delphi with an Indigenous panel

Background

There is significant disparity between the respiratory health of Indigenous and non-Indigenous Australian infants. There is no culturally accepted measure to collect respiratory health outcomes in Indigenous infants. The aim of this study was to gain end user and expert consensus on the most relevant and acceptable respiratory and birth measures for Indigenous infants at birth, between birth and 6 months, and at 6 months of age follow-up for use in a research trial.

Methods

A three round modified Delphi process was conducted from February 2018 to April 2019. Eight Indigenous panel members, and 18 Indigenous women participated. Items reached consensus if 7/8 (≥80%) panel members indicated the item was ‘very essential’. Qualitative responses by Indigenous women and the panel were used to modify the 6 months of age surveys.

Results

In total, 15 items for birth, 48 items from 1 to 6 months, and five potential questionnaires for use at 6 months of age were considered. Of those, 15 measures for birth were accepted, i.e., gestational age, birth weight, Neonatal Intensive Care Unit (NICU) admissions, length, head circumference, sex, Apgar score, substance use, cord blood gas values, labour, birth type, health of the mother, number people living in the home, education of mother and place of residence. Seventeen measures from 1-to 6 months of age were accepted, i.e., acute respiratory symptoms (7), general health items (2), health care utilisation (6), exposure to tobacco smoke (1), and breastfeeding status (1). Three questionnaires for use at 6 months of age were accepted, i.e., a shortened 33-item respiratory questionnaire, a clinical history survey and a developmental questionnaire.

Conclusions

In a modified Delphi process with an Indigenous panel, measures and items were proposed for use to assess respiratory, birth and health economic outcomes in Indigenous Australian infants between birth and 6 months of age. This initial step can be used to develop a set of relevant and acceptable measures to report respiratory illness and birth outcomes in community based Indigenous infants.

Quality of social and emotional wellbeing services for families of young Indigenous children attending primary care centers; a cross sectional analysis

BACKGROUND

The quality of social and emotional wellbeing services for Indigenous families of young children is not known, in many settings especially services provided by primary care centers.

METHODS

Our primary objective was to assess delivery of social and emotional wellbeing services to the families of young (3-11 months) and older (12-59 months) Indigenous children attending primary care centers. Our secondary objective was to assess if delivery differed by geographic location. Two thousand four hundred sixty-six client files from 109 primary care centers across Australia from 2012 to 2014 were analysed using logistic regression and generalised estimating equations.

RESULTS

The proportion of families receiving social and emotional wellbeing services ranged from 10.6% (102) (food security) to 74.7% (1216) (assessment of parent child interaction). Seventy one percent (71%, 126) of families received follow up care. Families of children aged 3-11 months (39.5%, 225) were more likely to receive social and emotional wellbeing services (advice about domestic environment, social support, housing condition, child stimulation) than families of children aged 12-59 months (30.0%, 487) (adjusted odds ratio [aOR] 1.68 95% CI 1.33 to 2.13). Remote area families (32.6%, 622) received similar services to rural (29.4%, 68) and urban families (44.0%, 22) (aOR 0.64 95% CI 0.29, 1.44).

CONCLUSIONS

The families of young Indigenous children appear to receive priority for social and emotional wellbeing care in Australian primary care centers, however many Indigenous families are not receiving services. Improvement in resourcing and support of social and emotional wellbeing services in primary care centers is needed.

Validity of the Aboriginal children's health and well-being measure: Aaniish Naa Gegii?

BACKGROUND

Aboriginal children experience challenges to their health and well-being, yet also have unique strengths. It has been difficult to accurately assess their health outcomes due to the lack of culturally relevant measures. The Aboriginal Children's Health and Well-Being Measure (ACHWM) was developed to address this gap. This paper describes the validity of the new measure.

METHODS

We recruited First Nations children from one First Nation reserve in Canada. Participants were asked to complete the ACHWM independently using a computer tablet. Participants also completed the PedsQL. The ACHWM total score and 4 Quadrant scores were expected to have a moderate correlation of between 0.4 and 0.6 with the parallel PedsQL total score, domains (scale scores), and summary scores.

RESULTS

Paired ACHWM and PedsQL scores were available for 48 participants. They had a mean age of 14.6 (range of 7 to 19) years and 60.4 % were girls. The Pearson's correlation between the total ACHWM score and a total PedsQL aggregate score was 0.52 (p = 0.0001). The correlations with the Physical Health Summary Scores and the Psychosocial Health Summary Scores were slightly lower range (r = 0.35 p = 0.016; and r = 0.51 p = 0.0002 respectively) and approached the expected range. The ACHWM Quadrant scores were moderately correlated with the parallel PedsQL domains ranging from r = 0.45 to r = 0.64 (p ≤ 0.001). The Spiritual Quadrant of the ACHWM did not have a parallel domain in the PedsQL.

CONCLUSIONS

These results establish the validity of the ACHWM. The children gave this measure an Ojibway name, Aaniish Naa Gegii, meaning "how are you?". This measure is now ready for implementation, and will contribute to a better understanding of the health of Aboriginal children.

Improving delivery of health care to Aboriginal and Torres Strait Islander children

AIM

To identify opportunities to improve health-care delivery for urban Aboriginal and Torres Strait Islander children requiring hospital admission and to determine their characteristics.

METHODS

We analysed all documentation of admissions of Aboriginal and/or Torres Strait Islander children to a tertiary paediatric hospital in 2010. We reviewed the medical records to determine whether the Aboriginal status of patients was known, whether Aboriginal and/or Torres Strait Islander children and their families were reviewed by Aboriginal staff during admission and whether basic health-care quality indicators were met, including documentation of anthropometry, ear examination findings, immunisation status and catch-up immunisation delivery.

RESULTS

In 2010, 543 (2%) patients admitted to the institution were identified as Aboriginal and/or Torres Strait Islander: 140/538 (26.0%) were from the first decile (most disadvantaged) on Socio-Economic Indexes for Areas index. Of all admitted children, 148/543 (27.3%) were referred to Aboriginal health professionals during admission, more when length of stay was greater than 7 days (61% vs. 23%, P < 0.001). There was documentation of weight in 533/543 (98.2%), ear examinations in 64/543 (11.8%), immunisations being not up to date in 126/543 (23%), catch-up immunisation given in 7/126 (5.6%), Aboriginal and/or Torres Strait Islander status in 8/543 (1.5%) medical and 1/543 (0.2%) nursing discharge summaries.

CONCLUSIONS

We have identified several opportunities to improve culturally appropriate health-care delivery for Aboriginal and Torres Strait Islander children admitted to hospital, including improved recognition of Aboriginal and/or Torres Strait Islander status of patients, improved access to Aboriginal health professionals and increased performance and documentation of basic anthropometry, ear examination and immunisation catch-up.

The national incidence and clinical picture of SLE in children in Australia – a report from the Australian Paediatric Surveillance Unit

Objectives

The objectives of this paper are to prospectively determine the incidence of paediatric systemic lupus erythematosus (pSLE) in Australia as well as describe the demographics, clinical presentation and one-year outcome.

Study design

Newly diagnosed cases of pSLE were ascertained prospectively from October 2009 to October 2011 through the Australian Paediatric Surveillance Unit (a national monthly surveillance scheme for notification of childhood rare diseases) as well as national subspecialty groups. Questionnaires were sent to notifying physicians at presentation and at one year.

Results

The annual incidence rate was 0.32 per 105 children aged less than 16 years. The incidence was significantly higher in children of Asian or Australian Aboriginal and Torres Strait Islander parents. Approximately one-third of children underwent a renal biopsy at presentation and 7% required dialysis initially although only one child had end-stage kidney disease (ESKD) at one-year follow-up.

Conclusion

The incidence of pSLE in Australia is comparable to that worldwide with a significantly higher incidence seen in children of Asian and Australian Aboriginal and Torres Strait Islander backgrounds. Renal involvement is common but progression to ESKD, at least in the short term, is rare.

Evaluation of paediatric nursing-sensitive outcomes in an Australian population using linked administrative hospital data

BACKGROUND

Research into nursing-sensitive outcomes using administrative health data has focussed on hospitalised adults. However, we developed algorithms for the identification of 13 paediatric nursing-sensitive outcomes, which we seek to examine for clinical utility. The aims were to determine the rates of paediatric nursing-sensitive outcomes in a Western Australian hospital and ascertain sociodemographic and clinical characteristics associated with a greater risk of developing nursing-sensitive outcomes in hospitalised children.

METHOD

A retrospective cohort study used linked administrative data of all Western Australian children ≤18 years admitted to the only tertiary paediatric hospital in Perth between 1999 and 2009. Rates per 1,000 hospital separations and per 10,000 patient days were calculated for the following nursing-sensitive outcomes: lower respiratory tract infection (LRTI), gastrointestinal (GI) infection, pneumonia, sepsis, arrest/shock/respiratory failure, central nervous system complication, central venous line infection, infectious disease, pressure ulcer, failure to rescue, surgical wound infection, physiologic/metabolic derangement, and postoperative cardiopulmonary complications. Poisson multiple regression models were fitted to estimate rate ratios (RR) and 95% confidence intervals (CI) for suspected risk factors.

RESULTS

Linked records of 129,719 hospital separations were analysed. Rates ranged from 0.5/1,000 for pressure ulcer to 14.0/1,000 hospital separations for GI infections. Age was significantly associated with the risk of a nursing-sensitive outcome: compared with adolescents, toddlers had greater risk of GI infection (RR 9.89; 95% CI 6.24, 15.69); infants had 7.74 times greater risk of LRTI (95% CI 5.11, 11.75), while neonates had lower risks for sepsis (RR 0.26; 95% CI 0.08, 0.90) and physiologic/metabolic derangement (RR 0.12; 95% CI 0.04, 0.35). The risk of surgical wound infection was 7.78 times greater (95% CI 5.10, 11.86) for emergency admissions than elective admissions.

CONCLUSIONS

Seven of the 13 defined nursing-sensitive outcomes occurred with sufficient frequency (>100 events over the 10 year study period) to be potentially useful for monitoring the quality of nursing care. These nursing-sensitive outcomes are: LRTI, GI infection, pneumonia, surgical wound infection, physiologic/metabolic derangement, sepsis and postoperative cardiopulmonary complications. When used for quality improvement or to benchmark with other agencies, data need to be adjusted for, or stratified by age and admission type, to ensure equitable comparisons.

Have the health gaps between Indigenous and non-Indigenous Australian children changed over time? Results from an Australian National Representative Longitudinal Study

The purpose of this study was to evaluate the changes of health gaps between Indigenous and non-Indigenous children over time and to explore critical factors that contribute to the changes. We employed data consisting of two cohorts of Australian children: infant (0/1 year) and children (4/5 years) that are part of the Longitudinal Study of Australian Children. Health outcomes were measured by physical outcome index (POI) and parent-rated health during 2004, 2006 and 2008. We used first-order autoregressive modelling to examine the longitudinal relationship between the changes in health outcomes and possible contributing risk factors. The results showed that the trends of POIs between Indigenous and non-Indigenous children were closing, while the gap of parent-rated health between the two populations persisted. We found that health outcomes (both POI and parent-rated health) at an earlier time point (t - 1) were significant predictors of the outcomes at the later time point (t). Carer's depression status, socio-economic position and neighbourhood liveability had significant and consistent impacts on parent-rated health, but had only varying impacts on POIs between Indigenous and non-Indigenous children as well as between the birth cohorts at different time periods. Similarly, low birth weight, carer's binge drinking behaviour and other risk factors showed such varying impacts at a particular time period. The study implied that appropriate interventions accompanied by monitoring of health outcomes are necessary in order to decrease the health gaps between Indigenous and non-Indigenous children.

Poor food and nutrient intake among Indigenous and non-Indigenous rural Australian children

BACKGROUND

The purpose of this study was to describe the food and nutrient intake of a population of rural Australian children particularly Indigenous children. Participants were aged 10 to 12 years, and living in areas of relative socio-economic disadvantage on the north coast of New South Wales.

METHODS

In this descriptive cross-sectional study 215 children with a mean age of 11.30 (SD 0.04) years (including 82 Indigenous children and 93 boys) completed three 24-hour food recalls (including 1 weekend day), over an average of two weeks in the Australian summer of late 2005.

RESULTS

A high proportion of children consumed less than the Australian Nutrient Reference Values for fibre (74-84% less than Adequate Intake (AI)), calcium (54-86% less than Estimated Average Requirement (EAR)), folate and magnesium (36% and 28% respectively less than EAR among girls), and the majority of children exceeded the upper limit for sodium (68-76% greater than Upper Limit (UL)). Energy-dense nutrient-poor (EDNP) food consumption contributed between 45% and 49% to energy. Hot chips, sugary drinks, high-fat processed meats, salty snacks and white bread were the highest contributors to key nutrients and sugary drinks were the greatest per capita contributor to daily food intake for all. Per capita intake differences were apparent by Indigenous status. Consumption of fruit and vegetables was low for all children. Indigenous boys had a higher intake of energy, macronutrients and sodium than non-Indigenous boys.

CONCLUSIONS

The nutrient intake and excessive EDNP food consumption levels of Australian rural children from disadvantaged areas are cause for concern regarding their future health and wellbeing, particularly for Indigenous boys. Targeted intervention strategies should address the high consumption of these foods.

Successful chronic disease care for Aboriginal Australians requires cultural competence

OBJECTIVE

To review the literature to determine the attributes of culturally appropriate healthcare to inform the design of chronic disease management (CDM) models for Aboriginal patients in urban general practice.

METHODS

A comprehensive conceptual framework, drawing on the Access to Care, Pathway to Care, Chronic Care, Level of Connectedness, and Cultural Security, Cultural Competency and Cultural Respect models, was developed to define the search strategy, inclusion criteria and appraisal methods for the literature review. Selected papers were reviewed in detail if they examined a chronic disease intervention for an Aboriginal population and reported on its evaluation, impacts or outcomes.

RESULTS

In the 173 papers examined, only 11 programs met the inclusion criteria. All were programs conducted in rural and remote Aboriginal community-controlled health services. Successful chronic disease care and interventions require adequate Aboriginal community engagement, utilising local knowledge, strong leadership, shared responsibilities, sustainable resources and integrated data and systems. These success factors fitted within the conceptual framework developed.

CONCLUSIONS

Research and development of culturally appropriate CDM models concurrently in both urban and rural settings will enable more rigorous evaluation, leading to stronger evidence for best practice. A partnership of mainstream and Aboriginal-controlled health services is essential to successfully 'close the gap'.

IMPLICATIONS

Findings will inform and guide the development, implementation and evaluation of culturally appropriate CDM in mainstream general practice and primary care.

Persistence of acute rheumatic fever in a tertiary children's hospital

AIM

To identify and describe all children admitted with acute rheumatic fever (ARF) to a tertiary paediatric hospital in Sydney over a 9-year period and to describe their demographic and clinical characteristics, management and short-term outcomes. Delays in diagnosis, recurrence of ARF and use of secondary prophylaxis were also documented.

METHODS

Retrospective review of medical records for children aged < 15 years admitted to the Children's Hospital at Westmead, Sydney, with ARF (International Classification of Diseases (ICD)-10 classification I0.0-109.9) during 2000-2008. Only cases meeting the National Heart Foundation of Australia diagnostic criteria for ARF were included.

RESULTS

Twenty-six children met the National Heart Foundation of Australia criteria for ARF. The median age was 11.5 years (range 5.8-14.6) and 15 (58%) were male. Ten (38%) identified as Pacific Islander, and 5 (19%) as Aboriginal and Torres Strait Islander (ATSI). Most (n= 20, 77%) lived in suburban Sydney, and 69% were classified in the two most disadvantaged quintiles on the Index of Relative Socioeconomic Disadvantage and Advantage. Four (15%) had Sydenham's chorea, and 81% had carditis (mitral and/or aortic regurgitation). Six (23%) children had previous ARF. Antibiotic prophylaxis to prevent recurrent ARF was prescribed in all cases, but 50% received oral penicillin, rather than by intramuscular injection. Barriers to timely diagnosis were identified in 81%, including delayed presentation and delayed referral.

CONCLUSION

Most children presenting to the hospital with ARF lived in disadvantaged areas of Sydney. Pacific Islander and Aboriginal and Torres Strait Islander children were over-represented. Mitigation of RHD requires early identification of ARF and appropriate delivery of secondary prophylaxis.

Acute and persistent diarrhea

Socially disadvantaged Indigenous infants and children living in western industrialized countries experience high rates of infectious diarrhea, no more so than Aboriginal children from remote and rural regions of Northern Australia. Diarrheal disease, poor nutrition, and intestinal enteropathy reflect household crowding, inadequate water and poor sanitation and hygiene. Acute episodes of watery diarrhea are often best managed by oral glucose-electrolyte solutions with continuation of breastfeeding and early reintroduction of feeding. Selective use of lactose-free milk formula, short-term zinc supplementation and antibiotics may be necessary for ill children with poor nutrition, persistent symptoms, or dysentery. Education, high standards of environmental hygiene, breastfeeding, and immunization with newly licensed rotavirus vaccines are all needed to reduce the unacceptably high burden of diarrheal disease encountered in young children from Indigenous communities.

History, law, and policy as a foundation for health care delivery for Australian indigenous children

This article identifies significant historical and contemporary issues, programs, and progress to better understand the current policy in Australia relating to Aboriginal child health and well-being. A legislative perspective gives context to contemporary issues based on legally sanctioned historical practices specifically designed to make Aboriginal peoples disappear, particularly through the control and assimilation of Indigenous children.

Undernutrition and obesity in indigenous children: epidemiology, prevention, and treatment

Over the past 50 years there has been a shift in nutritional problems amongst Indigenous children in developed countries from under-nutrition and growth faltering to overweight and obesity; the major exception is small numbers of Indigenous children predominately living in remote areas of Northern Australia. Nutritional problems reflect social disadvantage and occur with disproportionately high incidence in all disadvantaged subgroups. There is limited evidence of benefit from any strategies to prevent or treat undernutrition and obesity; there are a limited number of individual studies with generalizable high grade evidence of benefit. Potential solutions require a whole of society approach.

Evaluation of a national resource to identify and study rare diseases: the Australian Paediatric Surveillance Unit

AIMS

To evaluate the Australian Paediatric Surveillance Unit (APSU).

METHODS

We used criteria recommended by the Centres for Disease Control and Prevention (CDC) for evaluating surveillance systems and reviewed productivity, response rates, completeness of the mailing list and impacts of APSU studies. Anonymous evaluation questionnaires were sent to 1260 reporting clinicians, 42 researchers and 86 public health professionals to seek their feedback as users of the APSU.

RESULTS

APSU provides national epidemiological and clinical data about rare childhood conditions that inform public health policy and clinical practice. Between 2000 and 2007, APSU data were disseminated in 106 journal articles, 207 scientific presentations and 85 media items. Of paediatricians and paediatric sub-specialists actively practicing in Australia and listed as Fellows of the Royal Australasian College of Physicians, 92% participate in APSU surveillance. An average 96% of monthly report cards were returned per annum since 2000. Sensitivity of case ascertainment was difficult to calculate for many conditions because alternative sources of ascertainment do not exist. Evaluation questionnaires were returned by 818 (65%) reporting clinicians, who believe APSU is valuable for generating knowledge 665 (81%), identifying research priorities 635 (78%), guiding clinical practice 572 (70%) and informing public health policy 575 (70%). Similar responses were received from researchers and public health professionals.

CONCLUSIONS

The APSU fulfils its objectives and meets relevant CDC criteria for usefulness, simplicity, acceptability and representativeness, sensitivity and timeliness of data quality. However, stability is threatened by lack of continuing core funding. APSU is highly productive and valued by child health clinicians, researchers and public health professionals.

Indigenous child health research: a critical analysis of Australian studies

OBJECTIVES

To conduct a critical and systematic analysis of descriptive studies regarding the health, development and wellbeing status of Indigenous children in Australia and to map them according to 1) Reported Indigenous involvement in the research process; 2) Domains of the life-course model of health; and 3) Geographical location of the Indigenous child population sample.

METHODS

A search of electronic databases, targeted websites and reference lists of relevant papers. Studies from 1958 to 2005 with clear methods and results were included. Data were extracted, mapped and analysed according to domains of the life-course model of health and development, study location, and reported level of Indigenous involvement.

RESULTS

217 studies were eligible. Research predominantly addressed physical health (75.1%) with few studies addressing mental health and wellbeing (2.8%) or health determinants (27.6%). Indigenous involvement in the research process was not apparent in 71.4% of studies, although it appears to be increasing. Compared with 10.6% in metropolitan locations, 67.2% of the studies were conducted in very remote areas. Remaining studies were conducted in remote or regional areas or were national.

CONCLUSIONS

More work is needed to establish an evidence base of Australian Indigenous child health and wellbeing that is founded on Indigenous values, knowledge and participation. Not withstanding the significant need to address issues of core morbidity and physical health for Indigenous children, more research addressing emotional and social health and wellbeing is required, as are research questions of importance to Indigenous children living in urban settings.

Measuring the quality of hospital care: an inventory of indicators

BACKGROUND

Development of indicators to measure health-care quality has progressed rapidly. This development has, however, rarely occurred in a systematic fashion, and some aspects of care have received more attention than others. The aim of this study is to identify and classify indicators currently in use to measure the quality of care provided by hospitals, and to identify gaps in current measurement.

METHODS

A literature search was undertaken to identify indicator sets. Indicators were included if they related to hospital care and were clearly being collected and reported to an external body. A two-person independent review was undertaken to classify indicators according to aspects of care provision (structure, process or outcome), dimensions of quality (safety, effectiveness, efficiency, timeliness, patient-centredness and equity), and domain of application (hospital-wide, surgical and non-surgical clinical specialities).

RESULTS

383 discrete indicators were identified from 22 source organizations or projects. Of these, 27.2% were relevant hospital-wide, 26.1% to surgical patients and 46.7% to non-surgical specialities, departments or diseases. Cardiothoracic surgery, cardiology and mental health were the specialities with greatest coverage, while nine clinical specialities had fewer than three specific indicators. Processes of care were measured by 54.0% of indicators and outcomes by 38.9%. Safety and effectiveness were the domains most frequently represented, with relatively few indicators measuring the other dimensions.

CONCLUSION

Despite the large number of available indicators, significant gaps in measurement still exist. Development of indicators to address these gaps should be a priority. Work is also required to evaluate whether existing indicators measure what they purport to measure.