Population-based linkage of health records in Western Australia: development of a health services research linked database

Disparities between Aboriginal and non-Aboriginal perinatal mortality rates in Western Australia from 1980 to 2015

BACKGROUND

Perinatal mortality rates are typically higher in Aboriginal than non-Aboriginal populations of Australia.

OBJECTIVES

This study aimed to examine the pattern of stillbirth and neonatal mortality rate disparities over time in Western Australia, including an evaluation of these disparities across gestational age groupings.

METHODS

All singleton births (≥20 weeks gestation) in Western Australia between 1980 and 2015 were included. Linked data were obtained from core population health datasets of Western Australia. Stillbirth and neonatal mortality rates and percentage changes in the rates over time were calculated by Aboriginal status and gestational age categories.

RESULTS

From 1980 to 2015, data were available for 930 926 births (925 715 livebirths, 5211 stillbirths and 2476 neonatal deaths). Over the study period, there was a substantial reduction in both the Aboriginal (19.6%) and non-Aboriginal (32.3%) stillbirth rates. These reductions were evident in most gestational age categories among non-Aboriginal births and in Aboriginal term births. Concomitantly, neonatal mortality rates decreased in all gestational age windows for both populations, ranging from 32.1% to 77.5%. The overall stillbirth and neonatal mortality rate differences between Aboriginal and non-Aboriginal birth decreased by 0.6 per 1000 births and 3.9 per 1000 livebirths, respectively, although the rate ratios (RR 2.51, 95% CI 2.14, 2.94) and (RR 2.94, 95% CI 2.24, 3.85), respectively reflect a persistent excess of Aboriginal perinatal mortality across the study period.

CONCLUSIONS

Despite steady improvements in perinatal mortality rates in Western Australia over 3½ decades, the gap between Aboriginal and non-Aboriginal rates remains unchanged in relative terms. There is a continuing, pressing need to address modifiable risk factors for preventable early mortality in Aboriginal populations.

The Impact of Pneumococcal Vaccination on Bacterial and Viral Pneumonia in Western Australian Children: Record Linkage Cohort Study of 469589 Births, 1996-2012

Background

Pneumococcal conjugate vaccine (PCV) was included in Australia's National Immunisation Program for all children from 2005. We assessed the impact of PCV on all-cause and pathogen-specific pneumonia hospitalizations in Western Australian (WA) children aged ≤16 years.

Methods

All hospitalizations with pneumonia-related International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification diagnosis codes occurring in WA-born children (1996-2012) were linked to pathology records. Age-specific incidence rate ratios and temporal trends for all-cause and pathogen-specific pneumonia hospitalizations were calculated before and after PCV introduction.

Results

Among 469589 births, there were 15175 pneumonia-related hospitalizations. Hospitalization rates were 6.7 (95% confidence interval, 6.4-6.9) times higher in Aboriginal than in non-Aboriginal children. Following PCV introduction, all-cause pneumonia hospitalizations showed significant declines across all age groups. A pathogen was identified in 2785 of 6693 (41.6%) pneumonia hospitalizations that linked to a pathology record. Respiratory syncytial virus (RSV) was most frequently identified, with RSV-associated pneumonia hospitalization rates of 89.6/100000 child-years in Aboriginal and 26.6/100000 child-years in non-Aboriginal children. The most common bacterial pathogen was Streptococcus pneumoniae in Aboriginal children (32.9/100000 child-years) and Mycoplasma pneumoniae in non-Aboriginal children (8.4/100000 child-years). Viral pneumonia rates declined in all children following PCV introduction, with the greatest declines seen in non-Aboriginal children; declines in bacterial pneumonia were observed in non-Aboriginal children.

Conclusions

Based on our ecological analyses, PCV seems to have had an impact on hospitalizations for pneumonia, suggesting that the pneumococcus is likely to play a role in both bacterial and viral pneumonia. Respiratory viruses remain an important pathogen in childhood pneumonia. Vaccines targeting respiratory viruses are needed to combat the residual burden of childhood pneumonia.

Risk of Pelvic Inflammatory Disease in Relation to Chlamydia and Gonorrhea Testing, Repeat Testing, and Positivity: A Population-Based Cohort Study

Background

There is uncertainty around whether the risks of pelvic inflammatory disease (PID) differ following Chlamydia trachomatis (chlamydia) and Neisseria gonorrhoeae (gonorrhea) infection. We quantified the risk of PID associated with chlamydia and gonorrhea infection and subsequent repeat infections in a whole-population cohort.

Methods

A cohort of 315123 Western Australian women, born during 1974-1995, was probabilistically linked to chlamydia and gonorrhea testing records and to hospitalizations and emergency department presentations for PID from 2002 to 2013. Time-updated survival analysis was used to investigate the association between chlamydia and gonorrhea testing, and positivity, and risk of PID.

Results

Over 3199135 person-years, 120748 women had pathology test records for both chlamydia and gonorrhea, 10745 chlamydia only, and 653 gonorrhea only. Among those tested, 16778 (12.8%) had ≥1 positive chlamydia test, 3195 (2.6%) ≥1 positive gonorrhea test, and 1874 (1.6%) were positive for both. There were 4819 PID presentations (2222 hospitalizations, 2597 emergency presentations). Adjusting for age, Aboriginality, year of follow-up, health area, and socioeconomic status, compared to women negative for chlamydia and gonorrhea, the relative risk (adjusted incidence rate ratio) of PID was 4.29 (95% confidence interval [CI], 3.66-5.03) in women who were both chlamydia and gonorrhea positive; 4.54 (95% CI, 3.87-5.33) in those only gonorrhea positive; and 1.77 (95% CI, 1.61-1.94) in those only chlamydia positive.

Conclusions

Gonorrhea infection conferred a substantially higher risk than chlamydia of hospitalization or emergency department presentation for PID. The emergence of gonorrhea antimicrobial resistance may have a serious impact on rates of PID and its associated reproductive health sequelae.

Regularity of contact with GPs: Measurement approaches to improve valid associations with hospitalization

BACKGROUND

Studies examine longitudinal continuity of GP contact though few consider 'regularity of GP contact', i.e., the dispersion of contacts over time. Increased regularity may indicate planned ongoing care. Current measures of regularity may be correlated with the number of contacts and may not isolate the phenomenon of interest.

OBJECTIVES

To compare two published and one newly developed regularity index in terms of their ability to measure regularity of GP contacts independently of the number of contacts and the impact on their association with hospitalization.

METHODS

A cohort at risk of diabetes-related hospitalization in Western Australia from 1990 to 2004 was identified using linked administrative data. For each regularity index, relationships with number of GP contacts were assessed. Hospitalization was then regressed on each index with and without number of contacts as a covariate.

RESULTS

Among 153,414 patients the new regularity index showed a reduced association with number of contacts compared with existing indices. Associations with hospitalization differed between measures; for previously published indices, there were no significant associations between regularity and hospitalization, whereas on the new index, most regular GP contact was associated with reduced hospitalization (IRR = 0.90, 95% CI = 0.88-0.93). When number of contacts was added as a covariate, point estimates for this index showed little change, whereas for existing measures this addition changed point estimates.

CONCLUSION

A new measure of regularity of GP contact was less correlated with the number of contacts than previously published measures and better suited to estimating unconfounded relationships of regularity with hospitalization.

Familial and non-familial risk factors associated with colorectal cancer survival in young and middle-aged patients

BACKGROUND

Survival following colorectal cancer (CRC) survival may be influenced by a number of factors including family history, individual medical history, and comorbidities. The impact of these factors may vary based on the patient's age.

METHODS

The study cohort consisted of individuals born in Western Australia between 1945 and 1996, who had been diagnosed with CRC prior to 2015 (n = 3220). Hospital, cancer, and mortality data were extracted for each patient from state health records and were used to identify potential risk factors associated with CRC survival. Family linkage data, in combination with cancer registry data, were used to identify first-degree family members with a history of CRC. The association between survival following CRC diagnosis and identified risk factors was examined using Cox proportional hazard models.

RESULTS

Age and sex were not significantly associated with survival in young patients. However, in middle-aged patients increasing age (HR 1.03, 95% CI 1.01-1.05, p = 0.003) and being male (HR 0.72, 95% CI 0.60-0.87, p < 0.001) were associated with reduced survival. Being diagnosed with polyps and having a colonoscopy prior to CRC diagnosis were associated with improved survival in both young and middle-aged patients, while a history of non-CRC and liver disease was associated with reduced survival. In middle-aged patients, having diabetes-related hospital admissions (HR 1.53, 95% CI 1.15-2.03, p = 0.004) was associated with reduced survival.

CONCLUSIONS

In both young and middle-aged patients with CRC, factors associated with early screening and detection were associated with increased CRC survival while a history of liver disease and non-CRC was associated with decreased CRC survival.

Differential associations of ferritin and 25-hydroxyvitamin D with fasting glucose and diabetes risk in community dwelling older men

AIMS

We examined associations of ferritin and 25-hydroxyvitamin D with fasting glucose and prevalent diabetes in older men.

METHODS

Cross-sectional analysis of 4153 community-dwelling men aged 70 to 89 years in Western Australia. Plasma ferritin, 25-hydroxyvitamin D, and glucose were assayed. Diabetes was ascertained from self-report, medications, and fasting glucose.

RESULTS

There were 577 men with diabetes (13.9%). In the whole cohort, ferritin was associated with fasting glucose (0.051 mmol/L per 1 SD increase in ferritin, P = .006) and 25-hydroxyvitamin D was inversely associated (-0.085 mmol/L per 1 SD, P < .001). Ferritin was not associated with prevalent diabetes (highest vs. lowest quartile; >225 vs <66 μg/L: adjusted odds ratio [OR] 0.97, 95% confidence interval [CI], 0.74-1.27, P = .83). Higher vitamin D was associated with decreased odds of prevalent diabetes (highest vs lowest quartile; >82 nmol/L vs <53 nmol/L: OR = 0.57, 95% CI = 0.43-0.75, P < .001). There was no interaction between ferritin and vitamin D on diabetes risk.

CONCLUSIONS

In older men, ferritin is associated with fasting glucose but not prevalent diabetes. Higher 25-hydroxyvitamin D concentrations are independently associated with lower fasting glucose and reduced risk of diabetes. Clinical trials are required to determine whether interventions, which raise vitamin D concentrations, would reduce incidence of diabetes in this expanding demographic group.

The impact of pre-pregnancy body mass index and gestational weight gain on placental abruption risk: a systematic review and meta-analysis

PURPOSE

The aim of this systematic review was to evaluate the associations between pre-pregnancy body mass index and gestational weight gain and placental abruption.

METHODS

Relevant studies were identified from PubMed, EMBASE, Scopus and CINAHL. Unpublished findings from analyses of linked population-based data sets from Western Australia (2012-2015, n = 114,792) were also included. Studies evaluating pre-pregnancy body mass index and/or gestational weight gain and placental abruption were included. Two independent reviewers evaluated studies for inclusion and quality. Data including odds ratios (ORs) and 95% confidence intervals (CIs) were extracted and analysed by random effects meta-analysis.

RESULTS

21 studies were included, of which 15 were eligible for meta-analyses. The summary ORs for the association of being underweight, overweight and obese, and placental abruption, compared to normal weight women, were 1.4 (95% CI 1.1, 1.7), 0.8 (95% CI 0.8, 0.9) and 0.8 (95% CI 0.7, 0.9), respectively. These findings remained unchanged when each study was eliminated from the analysis and in subgroup analyses. Although data were scarce, women with gestational weight gain below the Institute of Medicine recommendations appeared to be at greater risk of abruption compared with women who had optimal weight gain.

CONCLUSIONS

Mothers that are underweight prior to or in early pregnancy are at a moderately increased risk of placental abruption.

Data Linkage: Canadian and Australian Perspectives on a Valuable Methodology for Intellectual and Developmental Disability Research

Data linkage holds great promise for generating new information about people with intellectual and developmental disabilities (IDD) as a population, yet few centers have developed the infrastructure to utilize this methodology. Two examples, from Canada and Australia, describe their efforts in building data linkage capabilities, and how linked databases can be used to identify persons with IDD and used for population-based research. The value of data linkage is illustrated through new estimates of prevalence of IDD; health service utilization patterns; associations with sociodemographic characteristics, and with physical and mental health conditions (e.g., chronic diseases, injury, fertility, and depression); and findings on equity in medical treatments. Examples are provided of findings used for governmental policy and program planning.

Institutional Variation in Quality of Cardiovascular Implantable Electronic Device Implantation: A Cohort Study

BACKGROUND

Cardiovascular implantable electronic devices (CIEDs) are associated with procedure-related complications, yet little is known about variation in complication rates among institutions that may suggest disparities in care quality.

OBJECTIVE

To assess institutional variation in risk-standardized complication rates (RSCRs) for CIED.

DESIGN

Cohort study.

SETTING

174 hospitals in Australia and New Zealand, 98 of which implanted at least 25 CIEDs during the study period.

PARTICIPANTS

81 304 patients older than 18 years (mean, 74.7 years [SD, 12.4]; 37.9% female) who received a new CIED (65 711 permanent pacemakers [PPMs] and 15 593 implantable cardioverter-defibrillators [ICDs]) in 2010 to 2015.

MEASUREMENTS

RSCRs and frequencies of major device-related complications during hospitalization or within 90 days of discharge.

RESULTS

Of the cohort, 6664 patients (8.2%) had a major complication. Although complication rates were higher for ICDs than PPMs (10.04% vs. 7.76%), 76.5% of all complications were attributable to PPMs (5098 vs. 1566 for ICDs). Among hospitals that implanted at least 25 CIEDs, the median RSCR was 8.1%; however, rates varied from 5.3% to 14.3%, with 22 hospitals identified as having RSCRs that differed significantly from the national average. Similar variation was observed when RSCRs for PPM implantation (n = 96 hospitals) (median RSCR, 7.6% [range, 5.4% to 12.9%]) were considered separately from those for ICD placement (n = 68 hospitals) (median RSCR, 9.7% [range, 6.2% to 16.9%]) and persisted when only elective procedures were assessed (n = 88 hospitals) (median RSCR, 7.4% [range, 4.7% to 13.0%]).

LIMITATION

Possible unmeasured confounding from the use of administrative data.

CONCLUSION

CIED complications are common and vary among hospitals, suggesting institutional variation in CIED care quality. Concerted clinical and policy interventions are needed to address CIED-related complications. These efforts should preferentially target PPMs, because most CIED complications are attributable to these devices.

PRIMARY FUNDING SOURCE

The Hospitals Contribution Fund Research Foundation.

Perinatal risk factors associated with skin infection hospitalisation in Western Australian Aboriginal and Non-Aboriginal children

BACKGROUND

Hospitalisation with skin infection in Western Australian (WA) Aboriginal children is common, with the highest rates in infants and children from remote WA.

OBJECTIVE

We aimed to quantify infant, maternal, and sociodemographic risk factors for skin infection hospitalisation in WA children, focussing on Aboriginal children aged <17 years.

METHODS

We conducted a retrospective population-based cohort study with linked perinatal and hospitalisation data on WA-born children (1996-2012), of whom 31 348 (6.7%) were Aboriginal. We used Cox regression to calculate adjusted hazard ratios and associated population attributable fractions (PAFs) for perinatal factors attributed to first hospitalisation with skin infection. To identify specific risk factors for early-onset infection, we further restricted the cohort to infants aged <1 year.

RESULTS

Overall, 5439 (17.4%) Aboriginal and 6750 (1.5%) non-Aboriginal children were hospitalised at least once with a skin infection. Aboriginal infants aged <1 year had the highest skin infection hospitalisation rate (63.2 per 1000 child-years). The strongest risk factors in Aboriginal children aged <17 years were socio-economic disadvantage, very remote location at birth, and multi-parity (≥3 previous pregnancies) accounting for 24%, 23%, and 15% of skin infection hospitalisations, respectively. Other risk factors included maternal age <20 years, maternal smoking during pregnancy, and low birthweight.

CONCLUSIONS

We have quantified the relative influence of perinatal risk factors associated with skin infection hospitalisations in WA children, providing measures indicating which factors have the potential to reduce the most hospitalisations. Our evidence not only supports existing calls for substantial government investment in addressing underlying social and environmental barriers to healthy skin in WA Aboriginal children but also identifies potential areas to target health promotion messaging at individuals/families on maternal smoking during pregnancy and skin hygiene for families.

Hospitalisations for maternal assault are associated with increased risk of child protection involvement

BACKGROUND

Previous research shows a co-occurrence between children's exposure to violence and child maltreatment.

OBJECTIVE

This study examined the risk of maltreatment allegations in children whose mothers had been hospitalised due to an assault.

PARTICIPANTS AND SETTING

The study used a retrospective cohort of children born in Western Australia between 1990-2009 (N = 524,534) using de-identified linked-administrative data.

METHODS

Multivariate Cox regression determined the adjusted and unadjusted hazard ratios for child maltreatment allegation in children with a mother hospitalised for assault. Models were adjusted for a range of sociodemographic characteristics.

RESULTS

One in five children had a maltreatment allegation following their mother's hospitalisation for assault. This increased to two in five children when the mother was assaulted in the prenatal period. Aboriginal children accounted for 57.6% of all allegations despite representing only 7.8% of the population. Children whose mother had a hospitalisation for assault were nine-times (HR = 9.20, 95%CI: 8.98-9.43) more likely to have a subsequent maltreatment allegation than children whose mother did not have a hospitalisation for assault. Following adjustment for confounding factors, both Aboriginal and non-Aboriginal children had an almost two-fold increased risk of maltreatment allegation (HR = 1.56, 95%CI: 1.43-1.70; HR = 1.93 95%CI:1.80-2.07).

CONCLUSIONS

Our study shows that child maltreatment allegation is common in children following a maternal hospitalisation for assault. Targeted early intervention is required for families with young children, and pregnant women experiencing violence. Importantly service staff need awareness of the impact of violence on families and the appropriate services to refer families to.

The pervasive effects of timing of parental mental health disorders on adolescent deliberate self-harm risk

Children whose parents have mental health disorders are at increased risk for deliberate self-harm (DSH). However, the effect of timing of parental mental health disorders on adolescent DSH risk remains under-researched. The aim of this study was to investigate how parental hospital admissions for mental health disorders and/or DSH in different developmental periods impact on the child’s DSH risk in adolescence. A nested case-control sample was compiled from a total population cohort sample drawn from administrative health records in Western Australia. The sample comprised 7,151 adolescents who had a DSH-related hospital admission (cases), and 143,020 matched controls who hadn’t had a DSH-related hospital admission. The occurrence of parental hospital admissions related to mental health disorders and/or DSH behaviours was then analysed for the cases and controls. The timing of the parental hospital admissions was partitioned into four stages in the child’s life course: (1) pre-pregnancy, (2) pregnancy and infancy, (3) childhood, and (4) adolescence. We found that adolescents of a parent with mental health and/or DSH-related hospital admissions in all developmental periods except pregnancy and infancy were significantly more likely than controls to have a DSH-related hospital admission. Compared to parental hospital admissions that occurred during childhood and adolescence, those that occurred before pregnancy conferred a higher risk for adolescent DSH: adjusted odds ratio (aOR) = 1.25 for having only one parent hospitalised and 1.66 for having both parents hospitalised for mental health disorders; aOR = 1.97 for having any parent hospitalised for DSH, all being significant at the level of p < .001. This study shows that timing is important for understanding intergenerational transmission of DSH risk. The pre-pregnancy period is as critical as period after childbirth for effective intervention targeting adult mental health disorders and DSH, highlighting the important role of adult mental health services in preventing DSH risk in future generations.

Trends in mental health related contacts among mothers of Aboriginal children in Western Australia (1990-2013): a linked data population-based cohort study of over 40 000 children

OBJECTIVE

This study examines the scale of maternal mental health related contacts among Australian Aboriginal children over time, and associations with socio-economic characteristics, geographical remoteness and maternal age.

DESIGN

A retrospective cohort study of the prevalence of maternal mental health related contacts among Aboriginal children born in Western Australia between 1990 and 2013.

SETTING

Population of Western Australia with de-identified linked administrative data from the Western Australian Department of Health.

PARTICIPANTS

All Aboriginal children born in Western Australia between 1990 and 2013 and their mothers.

PRIMARY OUTCOME MEASURE

Prevalence of maternal mental health related contacts among Aboriginal children born between 1990 and 2013. Mental health related contacts were identified using mental health related inpatient hospitalisations and outpatient contacts.

RESULTS

Almost 30% of cohort children were born to a mother with at least one mental health contact in the 5 years prior to birth, with 15% reported in the year prior to birth and the year post birth. There was a distinct increase in the prevalence of maternal mental health contacts between 1990 and 2013 (4-5% per year, with a peak in 2007). Maternal mental health contacts were associated with living in more disadvantaged areas and major cities, and having a mother aged over 20 years at birth.

CONCLUSIONS

The study affirms that mental health issues place a considerable burden on Aboriginal Australia, and suggests that many of the mental health issues that women develop earlier in life are chronic at the time of conception, during pregnancy and at birth. Early intervention and support for women in the earliest stages of family planning are required to alleviate the burden of mental health problems at birth and after birth. There is a clear need for policies on the development of a holistic healthcare model, with a multisector approach, offering culturally appropriate services for Aboriginal people.

Trends in chlamydia and gonorrhoea testing and positivity in Western Australian Aboriginal and non-Aboriginal women 2001-2013: a population-based cohort study

BACKGROUND

This study aimed to examine trends in chlamydia and gonorrhoea testing and positivity in Aboriginal and non-Aboriginal women of reproductive age.

METHODS

A cohort of 318002 women, born between 1974 and 1995, residing in Western Australia (WA) was determined from birth registrations and the 2014 electoral roll. This cohort was then probabilistically linked to all records of chlamydia and gonorrhoea nucleic acid amplification tests conducted by two large WA pathology laboratories between 1 January 2001 and 31 December 2013. Trends in chlamydia and gonorrhoea testing and positivity were investigated over time and stratified by Aboriginality and age group.

RESULTS

The proportion of women tested annually for chlamydia increased significantly between 2001 and 2013 from 24.5% to 36.6% in Aboriginal and 4.0% to 8.5% in non-Aboriginal women (both P-values <0.001). Concurrent testing was high (>80%) and so patterns of gonorrhoea testing were similar. Chlamydia and gonorrhoea positivity were substantially higher in Aboriginal compared with non-Aboriginal women; age-, region- and year-adjusted incidence rate ratios were 1.52 (95% confidence interval (CI) 1.50-1.69, P<0.001) and 11.80 (95% CI 10.77-12.91, P<0.001) respectively. Chlamydia positivity increased significantly in non-Aboriginal women aged 15-19 peaking in 2011 at 13.3% (95% CI 12.5-14.2%); trends were less consistent among 15-19-year-old Aboriginal women but positivity also peaked in 2011 at 18.5% (95% CI 16.9-20.2%). Gonorrhoea positivity was 9.7% (95% CI 9.3-10.1%), 6.7% (95% CI 6.4-7.0%), 4.7% (4.4-5.0%), and 3.1% (2.8-3.4%) among Aboriginal women aged respectively 15-19, 20-24, 25-29 and ≥30 years, compared with <1% in all age groups in non-Aboriginal women. Over time, gonorrhoea positivity declined in all age groups among Aboriginal and non-Aboriginal women.

CONCLUSION

Between 2001 and 2013 in WA chlamydia and gonorrhoea positivity remained highest in young Aboriginal women despite chlamydia positivity increasing among young non-Aboriginal women. More effective prevention strategies, particularly for young Aboriginal women, are needed to address these disparities.

Dementia complicating type 2 diabetes and the influence of premature mortality: the Fremantle Diabetes Study

AIMS

To investigate risk factors for, and the influence of premature mortality on, dementia complicating type 2 diabetes.

METHODS

Participants with type 2 diabetes in the community-based observational Fremantle Diabetes Study Phase 1 (n = 1291, mean age 64.0 years) were followed from 1993 to 1996 to end-June 2012. Incident dementia was identified from validated health databases. Dementia risk was assessed using Cox proportional hazards modelling supplemented by competing risk regression modelling in the total cohort and sub-groups defined by age of diabetes onset as mid-life (< 65 years) or late-life (≥ 65 years).

RESULTS

During mean ± SD follow-up of 12.7 ± 5.9 years, 717 participants (55.5%) died and 180 (13.9%) developed dementia. Overall, few risk factors predicted incident dementia and most predicted time to death. In mid-life diabetes, incident dementia was predicted by diabetes duration, cerebrovascular disease, schizophrenia, antipsychotic medication and the APOE ε4 allele. In late-life diabetes, risk factors were peripheral neuropathy, lack of exercise, lower fasting serum glucose, no antihypertensive therapy and the APOE ε4 allele. Competing risk analysis showed age to be a positive predictor compared with the inverse association in Cox models that suggested survivor bias in an older community-based cohort.

CONCLUSIONS

Dementia in type 2 diabetes is multifactorial. An association with diabetes duration, independent of most possible confounders, suggests that one or more unmeasured processes specific to diabetes may be implicated in the pathogenesis. The risk factors for dementia were also associated with an increased risk of death. This suggests that recently reported improvements in mortality in type 2 diabetes may be accompanied by reductions in dementia incidence.

Feature selection and transformation by machine learning reduce variable numbers and improve prediction for heart failure readmission or death

Background

The prediction of readmission or death after a hospital discharge for heart failure (HF) remains a major challenge. Modern healthcare systems, electronic health records, and machine learning (ML) techniques allow us to mine data to select the most significant variables (allowing for reduction in the number of variables) without compromising the performance of models used for prediction of readmission and death. Moreover, ML methods based on transformation of variables may potentially further improve the performance.

Objective

To use ML techniques to determine the most relevant and also transform variables for the prediction of 30-day readmission or death in HF patients.

Methods

We identified all Western Australian patients aged 65 years and above admitted for HF between 2003–2008 in linked administrative data. We evaluated variables associated with HF readmission or death using standard statistical and ML based selection techniques. We also tested the new variables produced by transformation of the original variables. We developed multi-layer perceptron prediction models and compared their predictive performance using metrics such as Area Under the receiver operating characteristic Curve (AUC), sensitivity and specificity.

Results

Following hospital discharge, the proportion of 30-day readmissions or death was 23.7% in our cohort of 10,757 HF patients. The prediction model developed by us using a smaller set of variables (n = 8) had comparable performance (AUC 0.62) to the traditional model (n = 47, AUC 0.62). Transformation of the original 47 variables further improved (p<0.001) the performance of the predictive model (AUC 0.66).

Conclusions

A small set of variables selected using ML matched the performance of the model that used the full set of 47 variables for predicting 30-day readmission or death in HF patients. Model performance can be further significantly improved by transforming the original variables using ML methods.

Infection Rates Before and After Diagnosis of IgA Vasculitis in Childhood: A Population-wide Study Using Non-exposed Matched Controls

OBJECTIVE

Clinical data suggest that infections can trigger IgA vasculitis (IgAV), but longterm observations are lacking. We compared rates, types, and microorganisms for serious infection before and after diagnosis for children with IgAV and non-exposed controls.

METHODS

Using population-based administrative linked health datasets we estimated incidence rates (IR) for serious infection per 1000 person-months for patients with IgAV (n = 504, age 5 yrs, 59.1% males) and controls matched for age, sex, and year of presentation (n = 1281, age 6 yrs, 66% males). Time zero (T0) was the date of IgAV diagnosis or equivalent date in controls, lookback (median 38 mos) was the period prior to T0, and followup (median 239 mos) was the period after T0.

RESULTS

During lookback, prevalence of serious infection was similar in patients with IgAV and controls (11.5% vs 9.5%, respectively), but patients with IgAV had a higher rate of upper respiratory tract infections [incidence rate ratio (IRR) 1.79; 95% CI 1.39-2.31] with shorter time between first serious infection and T0 (27 vs 43 mos; p = 0.02). During followup, patients were at a constant increased risk for serious infections (IRR 1.46, 95% CI 1.35-1.58). These rates were higher during followup: sepsis (IRR 12.6), pneumonia (IRR 6.19), upper respiratory tract infections (IRR 2.36), and skin infections (IRR 1.85). There was little overlap between patients with serious infections in the lookback and followup periods.

CONCLUSION

In patients with childhood IgAV there is an increased longterm risk for a broader spectrum of infections, which is unrelated to serious infections prior to diagnosis or treatment. This suggests disease-specific factors may have a lasting effect on immune competence in childhood IgAV.

Burn induced nervous system morbidity among burn and non-burn trauma patients compared with non-injured people

OBJECTIVE

Burns cause acute damage to the peripheral nervous system with published reports identifying that neurological changes after injury remain for a prolonged period. To shed some light on potential mechanisms, we assessed injury etiology and patterns of nervous system morbidity after injury by comparing long-term hospital admissions data of burns patients and other non-burn trauma patients with uninjured people.

METHODS

Linked hospital and death data of a burn patient cohort (n=30,997) in Western Australia during the period 1980-2012 were analysed along with two age and gender frequency matched comparison cohorts: non-burn trauma patients (n=28,647) and; non-injured people (n=123,399). The number of annual NS disease admissions and length of stay (LOS) were used as outcome measures. Multivariable negative binomial regression modelling was used to derive adjusted incidence rate ratios and 95% confidence intervals (IRR, 95% CI) and adjusted Cox regression models and hazard ratios (HR) were used to examine time to first nervous system admission after burn and incident admission rates.

RESULTS

The most common peripheral nervous system condition identified in each cohort (burn, non-burn trauma, uninjured) were episodic and paroxysmal disorders followed by nerve root and plexus disorders and polyneuropathies/peripheral NS conditions. Significantly elevated admission rates for NS conditions (IRR, 95% CI) were found for the burn (2.20, 1.86-2.61) and non-burn trauma (1.85, 1.51-2.27), compared to uninjured. Peripheral nervous system admission rates after injury (IRR, 95% CI) were significantly higher regardless of age at time of injury for the burn (<15years: 1.97, 1.49-2.61; 15-45: 2.70, 2.016-3.55; ≥45year: 1.62, 1.33-1.97) and non-burn trauma cohorts (<15years: 1.91, 1.55-2.35; 15-45: 1.94, 1.51-2.49; ≥45year: 1.42, 1.18-1.72), when compared to the uninjured. Significantly higher rates of incident NS hospitalisations were found for the burn cohort vs. uninjured cohort for a period of 15-years after discharge (0-5 years: HR, 95% CI: 1.97, 1.75-2.22; 5-15 years; HR, 95% CI: 1.44, 1.28-1.63). The non-burn trauma cohort had significantly higher incident nervous system admissions for 10 years after discharge (0-30 days: HR, 95% CI: 4.75, 2.44-9.23; 30days to 1-year HR, 95% CI: 2.95, 2.34-3.74; 1-5 years; HR, 95% CI: 1.47, 1.26-1.70; 5-10 years; HR, 95% CI: 1.34, 1.13-1.58).

CONCLUSIONS

Results suggest that injury patients are at increased risk of peripheral nervous system morbidity after discharge for a prolonged period of time. The time patterns associated with incident nervous system conditions suggest possible differences in underlying pathology and long-term patient care needs. Further research is needed to elucidate the underlying neuropathology.

A time-duration measure of continuity of care to optimise utilisation of primary health care: a threshold effects approach among people with diabetes

BACKGROUND

Literature highlighted the importance of timely access and ongoing care provided at primary care settings in reducing hospitalisation and health care resource uses. However, the effect of timely access to primary care has not been fully captured in most of the current continuity of care indices. This study aimed to develop a time-duration measure of continuity of primary care ("cover index") capturing the proportion of time an individual is under the potentially protective effect of primary health care contacts.

METHODS

An observational study was conducted on 36,667 individuals aged 45 years or older with diabetes mellitus extracted from Western Australian linked administrative data. Threshold effect models were used to determine the maximum time interval between general practitioner (GP) visits that afforded a protective effect against avoidable hospitalisation across complication cohorts. The optimal maximum time interval was used to compute a cover index for each individual. The cover was evaluated using descriptive statistics stratified by population socio-demographic characteristics.

RESULTS

The optimal maximum time between GP visits was 9-13 months for people with diabetes with no complication, 5-11 months for people with diabetes with 1-2 complications, and 4-9 months for people with diabetes with 3+ complications. The cover index was lowest among those aged 75+ years, males, Indigenous people, socio-economically disadvantaged and those in very remote areas.

CONCLUSIONS

This study developed a new measure of continuity of primary care that adds a time parameter to capturing longitudinal continuity. Cover has the potential to better capture underuse of primary care and will significantly contribute to the sparsely available methods for analysis of linked administrative data in evaluating continuity of care for people with chronic conditions.

Low birth weight and maternal incarceration in pregnancy: A longitudinal linked data study of Western Australian infants

Rationale

Improved birth weight outcomes have been reported for infants of mothers imprisoned during pregnancy relative to similarly disadvantaged mothers, however, findings are equivocal and evidence is lacking from jurisdictions outside the United States.

Objective

To investigate whether maternal imprisonment during pregnancy is a determinant of low birth weight (<2500 g) for Indigenous and non-Indigenous infants in Western Australia.

Methods

A longitudinal sample of 41,910 singleton infants born in Western Australia (October 1985-December 2013), was identified with linked administrative data and examined by five mutually exclusive categories of maternal corrections history; (i) imprisonment in pregnancy, (ii) imprisonment before pregnancy, (iii) first imprisonment after birth, (iv) community-based corrections record without imprisonment at any time, and (v) no corrections record at any time. Univariate and multivariate Poisson regression was performed to determine key risk factors for low birth weight. Prevalence of risk factors were calculated by maternal corrections history.

Results

After adjusting for other significant pregnancy risks, maternal imprisonment before (Indigenous RR 2.02, 95%CI 1.84–2.22, p<.001; non-Indigenous RR 2.48, 95%CI 1.98–3.12, p<.001) or during (Indigenous RR 1.96, 95%CI 1.68–2.29, p<.001; non-Indigenous RR 2.12, 95%CI 1.48–3.03, p<.001) pregnancy remained strong determinants of low birth weight, and carried greater risk than imprisonment after birth (Indigenous RR 1.58, 95%CI 1.44–1.74, p<.001; non-Indigenous RR 1.75, 95%CI 1.51–2.04, p<.001) or community-based corrections orders (Indigenous RR 1.32, 95%CI 1.21–1.43, p<.001; non-Indigenous RR 1.40, 95%CI 1.05–1.88, p<.001), relative to no corrections record. Pregnancy risk factors more prevalent amongst infants whose mothers were imprisoned before or during pregnancy included substance-use related service contacts, hospitalisation for injury, mental health service contacts, and having a sibling in contact with the child protection system.

Conclusion

Western Australian infants with mothers imprisoned before or during pregnancy experience elevated risk of low birth weight and exposure to maternal substance use, injury and mental distress in pregnancy.

Infant removals: The need to address the over-representation of Aboriginal infants and community concerns of another 'stolen generation'

OBJECTIVES

The removal of a child from their parents is traumatising, particularly in Aboriginal communities where a history of child removals has led to intergenerational trauma. This study will determine where disparities in child protection involvement exist among Aboriginal and non-Aboriginal children and characteristics associated with infant removals. Challenges faced by child protection and other agencies, and opportunities for overcoming these, are discussed.

METHODS

Data from both the Australian Institute of Health and Welfare and linked Western Australian government data was used to examine disparities between Aboriginal and non-Aboriginal children in the child protection and out-of-home care system.

RESULTS

Nationally, Aboriginal children are ten times more likely to be placed in out-of-home care than non-Aboriginal children and this disparity starts in infancy. Infants were removed from parents with high levels of risk. Aboriginal infants were at increased risk of being removed from women with substance-use problems and had greater proportions removed from remote, disadvantaged communities than were non-Aboriginal infants.

CONCLUSIONS

Aboriginal infants have a high rate of removal. Although there are many complexities to be understood and challenges to overcome, there are also potential strategies. The disparity between Aboriginal and non-Aboriginal infant removals needs to be seen as a priority requiring urgent action to prevent further intergenerational trauma.

Emergency health service contact and reincarceration after release from prison: A prospective cohort study

BACKGROUND

Adults released from prison often have complex health needs. They are at high risk of poor health outcomes and reincarceration, with health service use unlikely to be planned.

AIMS/HYPOTHESES

To determine the incidence of emergency health service (EHS) use, ambulance attendance and/or emergency department presentation, among 1,181 adults released from Australian prisons. We hypothesised that EHS contact would be associated with increased reincarceration risk.

METHODS

Baseline surveys were conducted within 6 weeks before release. Postrelease EHS contacts and reincarceration were identified through prospective data linkage. For each participant, EHS contacts within a 24-hour period were combined to make an episode. We used Cox proportional hazards regression to examine the relationship between EHS episodes and reincarceration, controlling for covariates.

RESULTS

More than half (53.3%) of participants had at least one EHS contact over a median of 25.6-month follow-up. In adjusted analyses, compared to those with no EHS contacts, the hazard of reincarceration was greater for participants who had one to three EHS episodes (hazard ratio [HR] = 1.84; 95% confidence interval [CI] [1.48, 2.29]) or four or more (HR = 2.35; 95% CI [1.67, 3.29]).

CONCLUSIONS/IMPLICATIONS FOR PRACTICE

Emergency department attendance by people with a history of imprisonment may be indicative of wider decompensation. Improved management of such patients may improve health outcomes and have collateral benefits for reducing reincarceration.

Incorporating geography into a new generalized theoretical and statistical framework addressing the modifiable areal unit problem

Background

All analyses of spatially aggregated data are vulnerable to the modifiable areal unit problem (MAUP), which describes the sensitivity of analytical results to the arbitrary choice of spatial aggregation unit at which data are measured. The MAUP is a serious problem endemic to analyses of spatially aggregated data in all scientific disciplines. However, the impact of the MAUP is rarely considered, perhaps partly because it is still widely considered to be unsolvable.

Results

It was originally understood that a solution to the MAUP should constitute a comprehensive statistical framework describing the regularities in estimates of association observed at different combinations of spatial scale and zonation. Additionally, it has been debated how such a solution should incorporate the geographical characteristics of areal units (e.g. shape, size, and configuration), and in particular whether this can be achieved in a purely mathematical framework (i.e. independent of areal units). We argue that the consideration of areal units must form part of a solution to the MAUP, since the MAUP only manifests in their presence. Thus, we present a theoretical and statistical framework that incorporates the characteristics of areal units by combining estimates obtained from different scales and zonations. We show that associations estimated at scales larger than a minimal geographical unit of analysis are systematically biased from a true minimal-level effect, with different zonations generating uniquely biased estimates. Therefore, it is fundamentally erroneous to infer conclusions based on data that are spatially aggregated beyond the minimal level. Instead, researchers should measure and display information, estimate effects, and infer conclusions at the smallest possible meaningful geographical scale. The framework we develop facilitates this.

Conclusions

The proposed framework represents a new minimum standard in the estimation of associations using spatially aggregated data, and a reference point against which previous findings and misconceptions related to the MAUP can be understood.

Electronic supplementary material

The online version of this article (10.1186/s12942-019-0170-3) contains supplementary material, which is available to authorized users.

Trends and predictors of extreme preterm birth: Western Australian population-based cohort study

Background

The preterm birth rate is rising in high-income countries and is associated with increased mortality and morbidity. Although the risks increase with greater prematurity and risk factors have been found to vary with gestational age and labour onset, few studies have focused on the myriad pathways to extreme preterm birth (20–27 weeks’ gestation). The current study investigated trends in extreme preterm birth by labour onset type and examined the antecedent risks to further our understanding around the identification of high-risk pregnancies.

Methods

Retrospective cohort study including all singleton extreme preterm births in Western Australia between 1986 and 2010. De-identified data from six core population health datasets were linked and used to ascertain extreme preterm births (excluding medical terminations and birth defects) after spontaneous onset of labour, preterm pre-labour rupture of membranes, and medically indicated labour onset. Trends over time in extreme preterm birth were analysed using linear regression. Multivariable regression techniques were used to assess the relative risks associated with each salient, independent risk factor and to calculate Population Attributable Risks (PARs).

Results

The extreme preterm birth rate including medical terminations and birth defects significantly increased over time whereas the extreme preterm birth rate excluding medical terminations and birth defects did not change. After medical terminations and birth defects were excluded, the rate of medically indicated extreme preterm births significantly increased over time whereas the rate of preterm pre-labour rupture of membranes extreme preterm births significantly reduced, and the rate of spontaneous extreme preterm births did not significantly change. In the multivariate analyses, factors associated with placental dysfunction accounted for >10% of the population attributable risk within each labour onset type.

Conclusions

First study to show that the increase in extreme preterm birth in high-income jurisdiction is no longer evident after medical terminations and birth defects are excluded. Interventions that identify and target women at risk of placental dysfunction presents the greatest opportunity to reduce extreme preterm births.

Harnessing the Power of Machine Learning in Dementia Informatics Research: Issues, Opportunities, and Challenges

Dementia is a chronic and degenerative condition affecting millions globally. The care of patients with dementia presents an ever-continuing challenge to healthcare systems in the 21st century. Medical and health sciences have generated unprecedented volumes of data related to health and wellbeing for patients with dementia due to advances in information technology, such as genetics, neuroimaging, cognitive assessment, free texts, routine electronic health records, etc. Making the best use of these diverse and strategic resources will lead to high-quality care of patients with dementia. As such, machine learning becomes a crucial factor in achieving this objective. The aim of this paper is to provide a state-of-the-art review of machine learning methods applied to health informatics for dementia care. We collate and review the existing scientific methodologies and identify the relevant issues and challenges when faced with big health data. Machine learning has demonstrated promising applications to neuroimaging data analysis for dementia care, while relatively less effort has been made to make use of integrated heterogeneous data via advanced machine learning approaches. We further indicate future potential and research directions in applying advanced machine learning, such as deep learning, to dementia informatics.

The cost-effectiveness of a telephone-based intervention to support caregivers of older people discharged from hospital

Background

A telephone intervention for caregivers of older people discharged from hospital was shown to improve preparedness to care, reduce caregiver strain and caregiver distress. No cost-effectiveness analysis has been published on this, or similar interventions. The study aims addressed here were to examine whether positive outcomes for caregivers resulting from the Further Enabling Care at Home (FECH) program changed the use and costs of health services by patients; and to assess cost-effectiveness.

Methods

A single-blind randomised controlled trial compared FECH to usual care. FECH involved a specially trained nurse addressing support needs of caregivers of older patients discharged from hospital. A minimum clinically important difference in preparedness to care was defined as an increase in Preparedness for Caregiving scale score of ≥ two points from baseline. Designated data collection was at: Time 1, within four days of discharge; Time 2, 15–21 days post-discharge; and Time 3, six weeks post-discharge. A last observation carried forward approach to loss to follow-up was used, with a sensitivity analysis including only those who completed all time points. Patient use of hospital, emergency department (ED) and ambulance services were captured for 12 weeks post-discharge using administrative data. Costs included nurse time supporting caregivers, resources used by the nurse, and time taken training the nurse to deliver FECH. Cost-effectiveness was assessed using decision trees for preparedness for caregiving.

Results

Sixty-two intervention dyads and 79 controls provided complete data. A significantly greater proportion of intervention group caregivers reported improved preparedness to care to Time 2 (36.4% v 20.9%, p = 0.029), though this was not sustained to Time 3. The intervention cost $AUD268.28 above usual care per caregiver. No significant differences were observed in health service use between groups. The incremental cost-effectiveness ratio for each additional caregiver reporting improved preparedness to care at Time 2 was $AUD1,730.84.

Conclusions

To our knowledge this is the first work to calculate the cost-effectiveness of a telephone-delivered intervention designed to support caregivers of older people post-discharge, and will support decision-making regarding implementation. Further research should examine different settings, and assess impacts on health service use with larger samples and a longer follow-up.

Trial registration

Australian and New Zealand Clinical Trial Registry: ACTRN12614001174673. Registered 07/11/2014.

Epidemiology of gastrostomy insertion for children and adolescents with intellectual disability

The largest group of recipients of pediatric gastrostomy have neurological impairment with intellectual disability (ID). This study investigated trends in first gastrostomy insertion according to markers of disadvantage and ID etiology. Linked administrative and health data collected over a 32-year study period (1983-2014) for children with ID born between 1983 and 2009 in Western Australia were examined. The annual incidence rate change over calendar year was calculated for all children and according to socioeconomic status, geographical remoteness, and Aboriginality. The most likely causes of ID were identified using available diagnosis codes in the linked data set. Of 11,729 children with ID, 325 (2.8%) received a first gastrostomy within the study period. The incidence rate was highest in the 0-2 age group and there was an increasing incidence trend with calendar time for each age group under 6 years of age. This rate change was greatest in children from the lowest socioeconomic status quintile, who lived in regional/remote areas or who were Aboriginal. The two largest identified groups of ID were genetically caused syndromes (15.1%) and neonatal encephalopathy (14.8%).Conclusion: Gastrostomy is increasingly used in multiple neurological conditions associated with ID, with no apparent accessibility barriers in terms of socioeconomic status, remoteness, or Aboriginality. What is Known: • The use of gastrostomy insertion in pediatrics is increasing and the most common recipients during childhood have neurological impairment, most of whom also have intellectual disability (ID). What is New: • Nearly 3% of children with ID had gastrostomy insertion performed, with the highest incidence in children under 3 years of age. • Gastrostomy use across different social groups was equitable in the Australian setting.

Temporal changes in the incidence and predictors of severe hypoglycaemia in type 2 diabetes: The Fremantle Diabetes Study

AIM

To determine the incidence of severe hypoglycaemia and its predictors in community-based patients with type 2 diabetes studied between 2008 and 2013 compared with those in a cohort of patients with type 2 diabetes from the same geographical area assessed a decade earlier.

METHODS

We studied 1551 participants (mean age 65.7 years, 51.9% men) with type 2 diabetes from the longitudinal observational Fremantle Diabetes Study Phase II (FDS2). Severe hypoglycaemia was ascertained as that requiring ambulance attendance, emergency department services and/or hospitalization. Cox proportional hazards modelling was used to determine predictors of a first episode of severe hypoglycaemia, and negative binomial regression was used to identify predictors of frequency.

RESULTS

Sixty-three participants (4.1%) experienced 83 episodes, representing an incidence of 1.34/100 participant-years (95% confidence interval [CI] 1.08 to 1.67; vs 1.67/100 participant-years [95% CI 1.31-2.13] in the Fremantle Diabetes Study Phase I [FDS1]; P = 0.18). Those experiencing severe hypoglycaemia experienced one to four episodes in both cohorts. The independent predictors of incident severe hypoglycaemia in the FDS2 were: older age; higher educational attainment; alcohol consumption; current smoking; sulphonylurea/insulin treatment; prior severe hypoglycaemia; renal impairment; and plasma N-terminal pro-B-type natriuretic peptide (NT-proBNP). The same variables except smoking were associated with frequency of severe hypoglycaemia. Most of these risk factors paralleled those in the FDS1, but current smoking and plasma NT-proBNP were novel.

CONCLUSIONS

The incidence and frequency of severe hypoglycaemia did not change between the Fremantle Diabetes Study phases but novel risk factors, including plasma NT-proBNP, were observed in the FDS2.

Machine learning-based prediction of heart failure readmission or death: implications of choosing the right model and the right metrics

AIMS

Machine learning (ML) is widely believed to be able to learn complex hidden interactions from the data and has the potential in predicting events such as heart failure (HF) readmission and death. Recent studies have revealed conflicting results likely due to failure to take into account the class imbalance problem commonly seen with medical data. We developed a new ML approach to predict 30 day HF readmission or death and compared the performance of this model with other commonly used prediction models.

METHODS AND RESULTS

We identified all Western Australian patients aged above 65 years admitted for HF between 2003 and 2008 in the linked Hospital Morbidity Data Collection. Taking into consideration the class imbalance problem, we developed a multi-layer perceptron (MLP)-based approach to predict 30 day HF readmission or death and compared the predictive performances using the performance metrics, that is, area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (AUPRC), sensitivity and specificity with other ML and regression models. Out of the 10 757 patients with HF, 23.6% were readmitted or died within 30 days of hospital discharge. We observed an AUC of 0.55, 0.53, 0.58, and 0.54 while an AUPRC of 0.39, 0.38, 0.46, and 0.38 for weighted random forest, weighted decision trees, logistic regression, and weighted support vector machines models, respectively. The MLP-based approach produced the highest AUC (0.62) and AUPRC (0.46) with 48% sensitivity and 70% specificity.

CONCLUSIONS

We show that for the medical data with class imbalance, the proposed MLP-based approach is superior to other ML and regression techniques for the prediction of 30 day HF readmission or death.

Geographical disparities in emergency department presentations for acute respiratory infections and risk factors for presenting: a population-based cohort study of Western Australian children

INTRODUCTION

Studies examining acute respiratory infections (ARIs) in emergency department (EDs), particularly in rural and remote areas, are rare. This study aimed to examine the burden of ARIs among Aboriginal and non-Aboriginal children presenting to Western Australian (WA) EDs from 2002 to 2012.

METHOD

Using a retrospective population-based cohort study linking ED records to birth and perinatal records, we examined presentation rates for metropolitan, rural and remote Aboriginal and non-Aboriginal children from 469 589 births. We used ED diagnosis information to categorise presentations into ARI groups and calculated age-specific rates. Negative binomial regression was used to investigate association between risk factors and frequency of ARI presentation.

RESULTS

Overall, 26% of presentations were for ARIs. For Aboriginal children, the highest rates were for those aged <12 months in the Great Southern (1233 per 1000 child-years) and Pilbara regions (1088 per 1000 child-years). Rates for non-Aboriginal children were highest in children <12 months in the Southwest and Kimberley (400 and 375 per 1000 child-years, respectively). Presentation rates for ARI in children from rural and remote WA significantly increased over time in all age groups <5 years. Risk factors for children presenting to ED with ARI were: male, prematurity, caesarean delivery and residence in the Kimberley region and lower socio-economic areas.

CONCLUSION

One in four ED presentations in WA children are for ARIs, representing a significant out-of-hospital burden with some evidence of geographical disparity. Planned linkages with hospital discharge and laboratory detection data will aid in assessing the sensitivity and specificity of ARI diagnoses in ED.

Perinatal Risk Factors Associated With Gastroenteritis Hospitalizations in Aboriginal and Non-Aboriginal Children in Western Australia (2000-2012): A Record Linkage Cohort Study

BACKGROUND

Gastroenteritis is a leading cause of childhood morbidity worldwide. We aimed to assess the maternal and infant characteristics and population attributable fractions associated with childhood gastroenteritis-related hospitalizations.

METHODS

We conducted a whole-of-population retrospective birth cohort study of 367,476 children live-born in Western Australia 2000-2012. We identified hospital admissions up to <15 years of age pertaining to these children, with a principal diagnosis code for infectious gastroenteritis. Cox regression was used to obtain the adjusted hazard ratios with 95% confidence intervals and the population attributable fractions associated with each risk factor in Aboriginal and non-Aboriginal children for their first gastroenteritis-related hospital admission.

RESULTS

There were a total of 15,888 gastroenteritis-related hospital admissions (25.7% occurring among non-Aboriginal children). The overall gastroenteritis hospitalization rate for children <15 years of age was 4.6/1000 child-years for non-Aboriginal children and 21.5/1000 child-years for Aboriginal children. Male gender, <20 years of maternal age, preterm birth, low birth weight, residence in remote regions of Western Australia and birth in the pre-rotavirus vaccine era were significant independent risk factors for gastroenteritis hospitalization in both Aboriginal and non-Aboriginal children. Additionally, birth by caesarean section and low socioeconomic status were identified as being associated with gastroenteritis hospitalization in non-Aboriginal children. Population attributable fractions suggest that 39% of all gastroenteritis hospitalizations in non-Aboriginal children (38% in Aboriginal children) could be averted if all children receive the rotavirus vaccine.

CONCLUSIONS

Given the beneficial effect of infant rotavirus vaccination in preventing all-cause gastroenteritis hospitalization, efforts should be taken to optimize rotavirus vaccine coverage in those at highest risk.

Use of the Rochester Epidemiology Project for Clinical Research in Colon and Rectal Surgery

The Rochester Epidemiology Project (REP), a longitudinal population-based database, is the largest epidemiologic database in the world. Originally established at Mayo Clinic in Rochester, Minnesota, the REP has been instrumental in defining the natural history of disease states and the impact of treatment in a well-defined population. In the United States, the REP has made important contributions to the colon and rectal literature, largely because healthcare is fragmented with no unique identifier to longitudinally follow-up with a patient throughout the system over his or her lifespan. Investigation with the REP has provided insight to the economic burden associated with inflammatory bowel disease, the benefit of screening for colorectal cancer, and the natural history of Mekel's and diverticular disease. In addition to practice changing research, the REP can be used as a model for future linkage systems in the United States.

The impact of first- and second-eye cataract surgery on motor vehicle crashes and associated costs

Objective

to evaluate the association between first- and second-eye cataract surgery and motor vehicle crashes for older drivers and the associated costs to the community.

Design

retrospective population-based cohort study.

Subjects

a total of 2,849 drivers aged 60 years and older who had undergone both first- and second-eye cataract surgery were involved in 3,113 motor vehicle crashes as drivers during the study period.

Methods

de-identified data were obtained using the Western Australian Data Linkage System from 1 January 2003 to 31 December 2015. Poisson regression analysis based on Generalised Estimating Equations was undertaken to compare the frequency of crashes in the year before first eye cataract surgery, between first and second eye surgery and 1 year after second eye surgery.

Results

first eye cataract surgery was associated with a significant 61% reduction in crash frequency (P < 0.001) and second eye surgery was associated with a significant 23% reduction in crashes (P < 0.001), compared to the year before first eye cataract surgery after accounting for age, gender, marital status, accessibility, socio-economic status, driving exposure and comorbidities. The estimated cost savings from the reduction in crashes in the year after second eye cataract surgery compared to the year before first eye cataract surgery was $14.9 million.

Conclusions

first- and second-eye cataract surgery were associated with a significant reduction in motor vehicle crashes, with first eye surgery having the greatest impact. These results provide encouragement for the timely provision of first- and second-eye cataract surgery for older drivers.

Development and Validation of a Simple Hip Fracture Risk Prediction Tool for Type 2 Diabetes: The Fremantle Diabetes Study Phase I

OBJECTIVE

To develop a type 2 diabetes hip fracture risk tool in community-based patients, to validate it in an independent cohort, and to compare its performance against the only published prediction equation to include type 2 diabetes as a risk factor (QFracture).

RESEARCH DESIGN AND METHODS

Hip fracture hospitalizations in 1,251 participants with type 2 diabetes aged 40-89 years from the longitudinal Fremantle Diabetes Study Phase I (FDS1) were ascertained between entry (1993-1996) and end-2012. Competing risk regression modeling determined independent predictors of time to first fracture over 10 years and the coefficients incorporated in a risk model. The model was validated in 286 participants with type 2 diabetes from the Busselton Health Study (BHS).

RESULTS

Fifty FDS1 participants (4.0%) experienced a first hip fracture during 10,306 person-years of follow-up. Independent predictors of fracture were older age, female sex, lower BMI, peripheral sensory neuropathy, and estimated glomerular filtration rate <45 mL/min/1.73 m². The model-predicted mean 10-year incident fracture risk was 3.3% with good discrimination, calibration, and accuracy. For a 3% cutoff, sensitivity was 76.0%, specificity 71.9%, positive predictive value (PPV) 10.1%, and negative predictive value (NPV) 98.6%. Model performance in the small BHS sample was also good (sensitivity 66.7%, specificity 79.8%, PPV 6.2%, and NPV 99.2%). QFracture performed well in FDS1 but required availability of 25 variables.

CONCLUSIONS

The FDS1 hip fracture risk equation is a simple validated adjunct to type 2 diabetes management that uses variables that are readily available in routine care.

Morbidity burden and community-based palliative care are associated with rates of hospital use by people with schizophrenia in the last year of life: A population-based matched cohort study

Objective

People with schizophrenia face an increased risk of premature death from chronic diseases and injury. This study describes the trajectory of acute care health service use in the last year of life for people with schizophrenia and how this varied with receipt of community-based specialist palliative care and morbidity burden.

Method

A population-based retrospective matched cohort study of people who died from 01/01/2009 to 31/12/2013 with and without schizophrenia in Western Australia. Hospital inpatient, emergency department, death and community-based care data collections were linked at the person level. Rates of emergency department presentations and hospital admissions over the last year of life were estimated.

Results

Of the 63508 decedents, 1196 (1.9%) had a lifetime history of schizophrenia. After adjusting for confounders and averaging over the last year of life there was no difference in the overall rate of ED presentation between decedents with schizophrenia and the matched cohort (HR 1.09; 95%CI 0.99–1.19). However, amongst the subset of decedents with cancer, choking or intentional self-harm recorded on their death certificate, those with schizophrenia presented to ED more often. Males with schizophrenia had the highest rates of emergency department use in the last year of life. Rates of hospital admission for decedents with schizophrenia were on average half (HR 0.53, 95%CI 0.44–0.65) that of the matched cohort although this varied by cause of death. Of all decedents with cancer, 27.5% of people with schizophrenia accessed community-based specialist palliative care compared to 40.4% of the matched cohort (p<0.001). Rates of hospital admissions for decedents with schizophrenia increased 50% (95% CI: 10%-110%) when enrolled in specialist palliative care.

Conclusion

In the last year of life, people with schizophrenia were less likely to be admitted to hospital and access community-based speciality palliative care, but more likely to attend emergency departments if male. Community-based specialist palliative care was associated with increased rates of hospital admissions.

Long-term mental health outcomes after unintentional burns sustained during childhood: a retrospective cohort study

Background

Burns are a devastating injury that can cause physical and psychological issues. Limited data exist on long-term mental health (MH) after unintentional burns sustained during childhood. This study assessed long-term MH admissions after paediatric burns.

Methods

This retrospective cohort study included all children (< 18 years) hospitalised for a first burn (n = 11,967) in Western Australia, 1980-2012, and a frequency matched uninjured comparison cohort (n = 46,548). Linked hospital, MH and death data were examined. Multivariable negative binomial regression modelling was used to generate incidence rate ratios (IRR) and 95% confidence intervals (CI).

Results

The burn cohort had a significantly higher adjusted rate of post-burn MH admissions compared to the uninjured cohort (IRR, 95% CI: 2.55, 2.07-3.15). Post-burn MH admission rates were twice as high for those younger than 5 years at index burn (IRR, 95% CI 2.06, 1.54-2.74), three times higher for those 5-9 years and 15-18 years (IRR, 95% CI: 3.21, 1.92-5.37 and 3.37, 2.13-5.33, respectively) and almost five times higher for those aged 10-14 (IRR, 95% CI: 4.90, 3.10-7.76), when compared with respective ages of uninjured children. The burn cohort had higher admission rates for mood and anxiety disorders (IRR, 95% CI: 2.79, 2.20-3.53), psychotic disorders (IRR, 95% CI: 2.82, 1.97-4.03) and mental and behavioural conditions relating to drug and alcohol abuse (IRR, 95% CI: 4.25, 3.39-5.32).

Conclusions

Ongoing MH support is indicated for paediatric burn patients for a prolonged period after discharge to potentially prevent psychiatric morbidity and associated academic, social and psychological issues.

Fit for purpose - developing a software platform to support the modern challenges of data linkage in Western Australia

The Western Australia Data Linkage System (WADLS) is maintained and operated by the WA Data Linkage Branch (DLB) at the Western Australian Department of Health. DLB has pioneered a number of data linkage innovations, including the facilitation of genealogical research via the Family Connections system and streamlined data delivery via the Custodian Administered Research Extract Server. DLB’s latest innovation is a new data linkage system called “DLS3”, which improves DLB’s capability and capacity to handle the increasing volume and complexity of its routine operations. DLS3 was built entirely in-house and customised to meet the specific challenges that DLB has encountered throughout over twenty years of experience with a wide variety of linkages. This article describes the development and rollout of DLS3, including its design, architecture, benefits and limitations.

A population-based comparison study of the mental health of patients with intentional and unintentional burns

Background

A number of studies report high prevalence of mental health conditions among burn patients. However there is a need to understand differences in the temporal relationship between mental health conditions and intentional and unintentional burns to hasten psychological prevention and intervention. This study aims to compare the socio-demographic profile, burn characteristics and pre- and post-burn psychiatric morbidity of burn patients by intent-of-injury.

Methods

De-identified linked hospital, death and mental health (MH) case registry data of burn patients hospitalised in Western Australia between 1 January 1980 and 30 June 2012 were analysed. Crude (observed) post-burn rates of mental health admissions were generated by burn intent-of-injury. Descriptive statistics were performed to compare the characteristics of the burn patients.

Results

A total of 30,997 individuals were hospitalised for a first burn; 360 (1.2%) had self-harm burns and 206 (0.7%) assault burns. Over the study period, admission rates for assault burns increased by 4.8% per year (95% confidence interval (CI) 3.1–6.5%) and self-harm burns increased 6.9% per year (95% CI 4.8–9.1%). Self-harm and assault burns occurred mainly among those aged 15 to 44 years (median age, interquartile range (IQR): self-harm 30 years, 22–40; assault 31 years, 23–38). Those with self-harm burns had a longer index hospital stay (median (IQR): self-harm 15 days (5–35) vs 4 days (1–11) assault vs 4 days (1–10) unintentional) and higher in-hospital mortality (7.2% self-harm vs 1.9% assault burns vs 0.8% unintentional). More than half (55.0%) of self-harm burns had a prior hospitalisation (5-year lookback) for a MH condition vs 10.7% of assault burns and 2.8% of unintentional burns. Crude post-burn rates of MH admissions per 100 person-years (PY) by intent-of-burn subgroups: self-harm 209 per 100 PY, assault burns 11 per 100 PY and unintentional burns 3 per 100 PY.

Conclusions

Intentional burn patients experienced significantly higher pre- and post-burn mental health morbidity along with significant adverse outcome in comparison with unintentional burns. Early psychological assessment and intervention could help in improving the MH of these patients.

Stillbirth in Western Australia, 2005-2013: the influence of maternal migration and ethnic origin

OBJECTIVE

To investigate prevalence rates and the risk of ante- and intrapartum stillbirth in Western Australia with respect to maternal country of birth and ethnic origin.

DESIGN, SETTING AND PARTICIPANTS

Whole population retrospective cohort analysis of de-identified, linked routinely collected birth, perinatal and mortality data for all births to non-Indigenous women in WA during 2005-2013.

MAIN OUTCOME MEASURES

Crude and adjusted odds ratios (aORs) with 95% confidence intervals were estimated by logistic regression and adjusted for confounding factors, for all stillbirths, antepartum stillbirths and intrapartum stillbirths, stratified by migrant status and ethnic background (white, Asian, Indian, African, Māori, other).

RESULTS

Women born overseas were more likely to have a stillbirth than Australian-born women (aOR, 1.26; 95% CI, 1.09-1.37). There was no significant difference for any type of stillbirth between Australian-born women of white and non-white backgrounds, but non-white migrant women were more likely than white migrants to have a stillbirth (OR, 1.42; 95% CI, 1.19-1.70). Compared with Australian-born women, migrants of Indian (aOR, 1.71; 95% CI, 1.17-2.47), African (aOR, 2.12; 95% CI, 1.46-3.08), and "other" ethnic origins (aOR, 1.43; 95% CI, 1.06-1.93) were more likely to have antepartum stillbirths; women of African (aOR, 5.08; 95% CI, 3.14-8.22) and "other" (aOR, 1.86; 95% CI, 1.15-3.00) background were more likely to have an intrapartum stillbirth.

CONCLUSIONS

Immigrants of African or Indian background appear to be at greater risk of ante- and intrapartum stillbirth in WA. Specific strategies are needed reduce the prevalence of stillbirth in these communities.

Evidence-based pharmacotherapies used in the postdischarge phase are associated with improved one-year survival in senior patients hospitalized with heart failure

AIM

Hospitalized heart failure (HF) patients have a poor prognosis postdischarge. We determined whether renin-angiotensin system inhibitors (RASI) and β-blockers dispensed to patients within 60 days post-HF hospital discharge are associated with improved 1-year survival.

METHODS

A retrospective population-based study was conducted in 4897 seniors, aged 65-84 years, alive at 60 days postindex HF hospitalization in Western Australia over 2003-2008. Dispensing of RASI and β-blocker dispensing was identified from the Pharmaceutical Benefits Scheme claims database linked to hospital admission and death records.

RESULTS

At 1-year posthospital discharge, the all-cause mortality and all-cause death or HF rehospitalization rate was 13.5% (n = 663) and 24.4% (n = 1193), respectively. Postdischarge RASI and β-blocker were dispensed in 77.4% and 53.0% of patients, respectively. Their use was associated with a lower inverse probability treatment weighted (IPTW) HR for 1-year mortality of 0.70, 95% CI 0.61-0.81 and 0.79, 95% CI 0.68-0.92, respectively (both P < 0.0001), with a survival advantage most evident in the subgroup (70.1%) of patients with ischemic HF. In the overall cohort, these therapies were also associated with reduced IPTW HRs for all-cause death or HF rehospitalization (both P < 0.005) but not for HF rehospitalization exclusively. Use of a β-blocker was associated with a reduced IPTW HR for HF rehospitalization in the ischemic HF subgroup only.

CONCLUSIONS

In a cohort of senior patients hospitalized with HF, dispensing of a RASI or β-blocker within 60 days postdischarge is associated with a 1-year survival benefit. Early postdischarge support programs after recent HF hospitalization should include measures to optimize adherence to evidence-based medications.

Operationalization and validation of a novel method to calculate adherence to polypharmacy with refill data from the Australian pharmaceutical benefits scheme (PBS) database

Background

Electronic health care data contain rich information on medicine use from which adherence can be estimated. Various measures developed with medication claims data called for transparency of the equations used, predominantly because they may overestimate adherence, and even more when used with multiple medications. We aimed to operationalize a novel calculation of adherence with polypharmacy, the daily polypharmacy possession ratio (DPPR), and validate it against the common measure of adherence, the medication possession ratio (MPR) and a modified version (MPR_m).

Methods

We used linked health data from the Australian Pharmaceutical Benefits Scheme and Western Australian hospital morbidity dataset and mortality register. We identified a strict study cohort from 16,185 patients aged ≥65 years hospitalized for myocardial infarction in 2003–2008 in Western Australia as an illustrative example. We applied iterative exclusion criteria to standardize the dispensing histories according to previous literature. A SAS program was developed to calculate the adherence measures accounting for various drug parameters.

Results

The study cohort was 348 incident patients (mean age 74.6±6.8 years; 69% male) with an admission for myocardial infarction who had cardiovascular medications over a median of 727 days (range 74 to 3,798 days) prior to readmission. There were statins (96.8%), angiotensin converting enzyme inhibitors (88.8%), beta-blockers (85.6%), and angiotensin receptor blockers (13.2%) dispensed. As expected, observed adherence values were higher with mean MPR (median 89.2%; Q₁: 73.3%; Q_3: 104.6%) than mean MPR_m (median 82.8%; Q₁: 68.5%; Q_3: 95.9%). DPPR values were the most narrow (median 83.8%; Q₁: 70.9%; Q_3: 96.4%). Mean MPR and DPPR yielded very close possession values for 37.9% of the patients. Values were similar in patients with longer observation windows. When the traditional threshold of 80% was applied to mean MPR and DPPR values to signify the threshold for good adherence, 11.6% of patients were classified as good adherers with the mean MPR relative to the DPPR.

Conclusion

In the absence of transparent and standardized equations to calculate adherence to polypharmacy from refill databases, the novel DPPR algorithm represents a valid and robust method to estimate medication possession for multi-medication regimens.

A prospective cohort examination of haematological parameters in relation to cancer death and incidence: the Busselton Health Study

BACKGROUND

Cancer risk is associated with serum iron levels. The aim of this study was to evaluate whether haematological parameters reflect serum iron levels and may also be associated with cancer risk.

METHODS

We studied 1564 men and 1769 women who were enrolled in the Busselton Health Study, Western Australia. Haematological parameters evaluated included haemoglobin (Hb), mean cell volume (MCV), mean cell haemoglobin (MCH) and mean cell haemoglobin concentration (MCHC) and red cell distribution width (RCDW). Statistical analyses included t-tests for quantitative variables, chi-square tests for categorical variables and Cox proportional hazards regression modelling for cancer incidence and death.

RESULTS

There was marginal evidence of an association between MCV (as a continuous variable) and non-skin cancer incidence in women (HR 1.15, 95% CI 1.013, 1.302; p = 0.030) but the hazard ratio was attenuated to non-significance after adjustment for serum ferritin (SF), iron and transferrin saturation (TS) (HR 1.11, 95% CI 0.972, 1.264; p = 0.126). There was strong evidence of an association between MCHC and prostate cancer incidence in men; the estimated hazard ratio for an increase of one SD (0.5) in MCHC was 1.27 (95% CI 1.064, 1.507; p = 0.008). These results remained significant after further adjustment for SF and iron; the estimated hazard ratio for an increase of one SD (0.5) in MCHC was 1.25 (p = 0.014, 95% CI 1.05 to 1.48).

CONCLUSIONS

The MCHC and MCV were associated with cancer incidence in a Western Australian population, although only MCHC remained associated with prostate cancer after adjusting with serum iron and TS (circulating iron) and SF (storage iron). Haematological parameters are thus of limited utility in population profiling for future cancer risk.

The relationship between circulating adiponectin, ADIPOQ variants and incident cardiovascular disease in type 2 diabetes: The Fremantle Diabetes Study

AIMS

To investigate the relationship between serum adiponectin, ADIPOQ variants and haplotypes, and cardiovascular disease (CVD) in type 2 diabetes (T2D).

METHODS

Baseline data including serum total adiponectin and 21 ADIPOQ polymorphisms were available for 1076 participants (mean age 64.0 years, 49.4% males) in a community-based cohort followed for an average of 12 years.

RESULTS

During 8843 patient-years of follow-up for coronary heart disease (CHD), 13,494 patient-years for ischaemic stroke (IS) and 12,028 patient-years for heart failure (HF), 40.4%, 11.8% and 31.9% of patients experienced a first episode of CHD, IS or HF, respectively. In Cox regression after adjustment for the most parsimonious models, log_e(serum adiponectin) and the ADIPOQ variant rs12495941 were inversely associated with incident CHD (hazard ratio [95% confidence interval] 0.79 [0.65-0.98] and 0.64 [0.44-0.94], respectively), while rs1648707 was positively associated with incident IS (2.05 [1.37-3.06]; all P ≤ 0.028). In males, rs9860747 and rs17366568 predicted CHD (0.22 [0.05-0.92] and 1.50 [1.01-2.20]; P ≤ 0.042), while rs1648707 and rs1063537 predicted IS (2.36 [1.32-4.23] and 2.09 [1.17-3.72]; P ≤ 0.012). In females, rs10937273 predicted CHD via an interaction with serum adiponectin (0.43 [0.21-0.91]; P = 0.027), while rs864265 predicted IS (0.43 [0.21-0.88], P = 0.021). The associations between ADIPOQ variants and outcomes were supported by haplotype block analysis. Neither serum adiponectin nor ADIPOQ variants predicted HF.

CONCLUSIONS

Serum total adiponectin and gender-specific ADIPOQ variants predict CHD and IS, but not HF, independently of other risk factors in community-based patients with T2D. In contrast to some previous studies, there was no relationship between a high serum total adiponectin and CVD.