Methodology Report

Estimating Childhood Homelessness Deaths in Australia

A Multi-Pathway Mortality Model for Quantifying Hidden Deaths Among Housing-Insecure Children

Research Report | January 2026 | Version 1.0

Key Finding

Our multi-pathway mortality model estimates 150–170 childhood deaths annually in Australia are attributable to homelessness and housing instability. This represents approximately 4–4.5 times the official count derived from administrative data linkage, which captures only 23–30% of actual deaths.

169

Our Central Estimate (Deaths/Year)

1. Introduction

1.1 The Problem of Invisible Deaths

In Australia, the deaths of homeless children are systematically undercounted. When a child dies, their housing status is rarely recorded in mortality data. The Australian Institute of Health and Welfare (AIHW) can only identify deaths linked to Specialist Homelessness Services (SHS) through complex administrative data linkage—and even this captures only a fraction of the true toll.

"The work of services and researchers is hindered partly by the fact that housing or homelessness status is not, or at least not accurately, captured in typical mortality data sources such as government registries of deaths or coronial records."— Tuson et al. (2024)

This creates a fundamental measurement problem: we cannot accurately count deaths that the data system was never designed to capture. Children who die while couch surfing, living in overcrowded housing, or whose families avoided services due to fear of child protection intervention simply do not appear in official statistics.

1.2 Purpose of This Model

This methodology presents a multi-pathway mortality model designed to estimate the true number of childhood deaths attributable to homelessness and housing instability in Australia. Rather than applying a single risk multiplier, we estimate deaths through distinct causal pathways, each with different:

• Age distributions and risk mechanisms

• Degrees of visibility to administrative data systems

• Evidence bases and uncertainty levels

1.3 Scope and Definitions

Age Range

This model covers children aged 0–17 years. We use age stratification because mortality risk varies significantly across childhood, and different pathways affect different age groups.

Definition of Homelessness

We adopt a broad definition encompassing:

• Primary homelessness: Rough sleeping, living in improvised dwellings

• Secondary homelessness: Crisis accommodation, couch surfing, temporary arrangements

• Tertiary homelessness: Severe overcrowding (4+ additional bedrooms needed)

• Housing instability: Frequent moves, imminent risk of homelessness, service-avoidant families

2. Methodology

2.1 Conceptual Framework

Our approach recognizes that deaths attributable to childhood homelessness fall into two categories:

Visible deaths: Those captured by AIHW through linkage of SHS records to the National Death Index. These represent the minimum confirmed count but are limited by:

• Data linkage failure (~21% of SHS records cannot be linked)

• Requirement that families accessed SHS within 12 months of death

• Correct recording of Statistical Linkage Key (SLK) data

Hidden deaths: Those occurring in populations invisible to administrative systems, including:

• Children in families who never access SHS due to fear, stigma, or barriers

• Young people couch surfing without service contact

• Deaths occurring >12 months after last SHS contact

• Deaths where housing status is not recognized or recorded

2.2 Data Sources

2.3 Establishing the Visible Baseline

From the AIHW NACS linked dataset (2012-13 to 2022-23):

"Over the study period, among the 14,000 people who received SHS support in the last year of life, around 1 in 77 (1.3%) were children aged 0–14 years (180 deaths)."— AIHW (2024)

Annualized visible deaths (0-14): 180 ÷ 10 = 18 deaths/year

For ages 15-17, we extrapolate from the under-25 data (6.2% of 14,000 = 868 deaths over 10 years). Assuming 15-17 represents ~30% of the 15-24 band:

Annualized visible deaths (15-17): (868 - 180) × 0.3 ÷ 10 ≈ 21 deaths/year

Total visible child deaths (0-17): 39 deaths/year

Data Linkage Adjustment

The AIHW acknowledges significant linkage limitations:

"A total of 1,262,977 records were linked from the Specialist Homelessness Service Collection cohort, which accounts for 79.25% of the 1,593,584 cohort SLKs."— AIHW (2024)

With a linkage success rate of ~79%, and assuming child records have slightly higher failure rates (estimated 75% success), we apply a linkage adjustment:

Linkage-adjusted visible deaths: 39 ÷ 0.75 = 52 deaths/year

2.4 Estimating the Hidden Homeless Child Population

Census Baseline

The ABS Census 2021 identified approximately 29,000 children aged 0-18 experiencing homelessness on Census night. However, this significantly undercounts several populations:

Category A: Couch Surfing (Secondary Homelessness)

"The population of young couch surfers can be underestimated when a person filling out the Census form on behalf of the young person does not know that they are unable to return home or do not have a home."— AIHW, Australia's Youth: Homelessness and Overcrowding

"In Australia, more than half of 15-to-24 year olds seeking housing assistance in 2019–2020 indicated that they were couchsurfing."— Green (2024), Geography Compass

We apply a 3× multiplier to census couch-surfing counts for children, reflecting systematic undercount.

Category B: Severe Overcrowding

"The majority (62% or 12,000) of children experiencing homelessness were living in severely overcrowded dwellings."— AIHW, Australia's Children: Homelessness

We estimate an additional 50% beyond census-captured overcrowding due to unreported arrangements.

Category C: Service-Avoidant Families

This represents perhaps the most significant hidden population:

"Service providers reported significant fear from clients about child protection involvement because of housing insecurity. Families were anxious children would be removed because they are going without food and essential items, aren't regularly attending school, and have significant exposure to violence and harm."— Centre for Excellence in Child and Family Welfare (2025)

"Several said that they stayed in abusive relationships because they were afraid the department would remove their children if they sought help. Some women avoided seeking medical assistance after incidents of domestic violence because they were afraid they would lose their children."— Human Rights Watch (2025)

We estimate a service-avoidant population equal in size to the known SHS child client base (~76,000).

Total Hidden Population Calculation

Total homeless/housing-insecure children: 29,000 + 118,000 = 147,000

This represents approximately 2.5% of Australia's ~5.8 million children aged 0-17.

2.5 Age-Stratified Risk Multipliers

Child mortality risk varies dramatically by age and cause. We derive homelessness risk multipliers from AIHW data and research evidence:

"Annually, the death rate of SHS clients was up to 1.7 times the rate of non-SHS clients; around 1.7–2.3 times for males and 1.0–1.4 times for females."— AIHW (2024)

"A Queensland study found that between 1990 and 2009, the suicide rate among homeless individuals was 27.6 per 100,000 – almost double that of those who were not homeless."— Arnautovska (2014)

3. The Multi-Pathway Mortality Model

3.1 Model Structure

Total estimated deaths are calculated as the sum of visible (linkage-adjusted) deaths plus deaths across multiple hidden pathways:

E total = E visible-adj + ESUDI + E suicide/OD + E FDV + E accident + E unlinked

3.2 Formal Expression

Complete Model Formula:

E_total = (E_AIHW / L_rate) + Σ(P_hidden,a × R_a × M_a) + (E_FDV × H_fraction) + (E_visible × U_rate) Where: E_AIHW = AIHW-reported linked deaths (39/year for children 0-17) L_rate = Linkage success rate (0.75) P_hidden,a = Hidden homeless population in age band a R_a = Baseline mortality rate for age band a (per capita) M_a = Homelessness risk multiplier for age band a E_FDV = Annual child FDV homicides (~20) H_fraction = Housing-attributable fraction (0.50) U_rate = Unlinked death rate (0.15)

3.3 Pathway Calculations

Pathway 1: Visible Deaths (Linkage-Adjusted)

E visible-adj = 39 ÷ 0.75 = 52 deaths/year

Pathway 2: SUDI in Hidden Homeless Infants

"In 2022, there were 117 SUDI deaths across Australia."— Red Nose (2022)

"The rates of SUDI deaths are three times as high in Aboriginal and Torres Strait Islander communities than in other communities."— NCBI (2018)

Calculation:

• Total SUDI deaths/year: 117

• Proportion in disadvantaged populations: 40%

• Housing-instability-attributable fraction: 30%

E SUDI = 117 × 0.40 × 0.30 = ~17 deaths/year

Pathway 3: Adolescent Suicide/Overdose (Hidden Population)

"520 children who have sought assistance from specialist homelessness services died between 2013-23, with suicide being the leading cause of death among those aged 12-17."— AIHW (2024)

For hidden homeless adolescents (12-17) without service contact:

• Hidden homeless adolescents: ~45,000

• Base mortality rate (15-17): 35/100,000

• Risk multiplier (hidden): 6.0×

• Cause-specific fraction (suicide/OD): 60%

E suicide/OD = 45,000 × (35/100,000) × 6.0 × 0.60 = ~57 deaths/year

Pathway 4: FDV-Related Deaths (Housing Barrier)

"Many victims (7%) and offenders (5%) had been involved with housing services before the murder, and more victims had engaged with a housing service before their death (7%) than with a specialist family violence service (5%)."— Fitz-Gibbon et al. (2024)

"Being unable to secure housing is credited with DFV victim-survivors continuing to live with the perpetrator."— Warren & McAuliffe (2021)

E FDV = 20 × 0.50 = ~10 deaths/year

Pathway 5: Accident/Illness (Hidden Children 1-11)

Applying age-specific rates with housing-instability multipliers:

Pathway 6: Unlinked Deaths

"More than three-quarters (77 per cent) of unaccompanied children who were homeless when support started were still homeless when the supports ended."— AIHW (2024)

Deaths occurring >12 months after last SHS contact (estimated 15% of visible deaths):

E unlinked = 52 × 0.15 = ~8 deaths/year

3.4 Total Estimate

Central Estimate Assembly:

3.5 Sensitivity Analysis

Running the model with different parameter assumptions:

95% Confidence Interval: 89 – 251 deaths/year

3.6 Simplified Formula for Ongoing Monitoring

For practical application:

Simplified Model

E_estimated = (E_AIHW × 1.33) + (P_hidden × R̄ × M̄) Where: E_AIHW = Latest AIHW annualized child deaths (~39) 1.33 = Linkage adjustment factor (1/0.75) P_hidden = Estimated hidden homeless children (~118,000) R̄ = Weighted average base mortality rate (~18/100,000) M̄ = Weighted average risk multiplier (~4.5) E_estimated = (39 × 1.33) + (118,000 × 0.00018 × 4.5) = 52 + 95 = 147 Plus FDV and unlinked adjustments (~18) = ~165 deaths/year

4. Model Validation

4.1 Cross-Validation Tests

Test 1: Ratio to Adult Deaths

AIHW reports ~1,400-1,500 adult SHS-connected deaths annually. If children represent ~10-12% of the homeless population but have lower base mortality rates, child deaths should be 8-15% of adult deaths.

Our estimate: 169 ÷ 1,450 = 11.7% ✓ Consistent

Test 2: Proportion of Total Child Mortality

Australia records ~1,700 child deaths annually. Our estimate suggests 169 are homelessness-related = 9.9%. Given 2.5% of children experience housing insecurity with 3-6× mortality elevation, this is plausible. ✓

Test 3: Coronial Data Comparison

Victorian coronial data shows ~100 child deaths annually requiring investigation, with ~15-20% involving socioeconomic distress. Extrapolating nationally: (100 × 0.175) ÷ 0.25 = 70 minimum. Our estimate exceeds this floor, as expected for a comprehensive model. ✓

5. Limitations and Caveats

Important Limitations

• Hidden population estimates rely on multipliers derived from service data and academic research, not direct observation

• Risk multipliers are partially extrapolated from adult data where child-specific data is unavailable

• Cause-of-death attribution is inherently uncertain for hidden populations

• Year-to-year variation may be significant but cannot be captured in this aggregate model

• Geographic variation (urban/rural, state differences) is not modelled separately

Despite these limitations, the fundamental finding—that official statistics capture only a fraction of actual deaths—is robust across all reasonable parameter ranges.

6. Data Sources and References

Primary Sources

1.Australian Institute of Health and Welfare (2024). People receiving specialist homelessness services support in last year of life. AIHW, Canberra. https://www.aihw.gov.au

2.Australian Institute of Health and Welfare (2024). Specialist Homelessness Services Annual Report 2023-24. AIHW, Canberra.

3.Australian Bureau of Statistics (2022). Census of Population and Housing: Estimating Homelessness, 2021. ABS Cat. 2049.0.

4.Red Nose Australia (2022). Media Fast Facts: SUDI Statistics. https://rednose.org.au

5.Homelessness Australia (2024). Child Homelessness Snapshot. https://homelessnessaustralia.org.au

6.Tuson, M. et al. (2024). "Tracking deaths of people who have experienced homelessness: a dynamic cohort study in an Australian city." BMJ Open, 14(3), e081260.

7.Brackertz, N. (2020). The role of housing insecurity and homelessness in suicidal behaviour. AHURI Evidence Check for the National Suicide Prevention Adviser.

8.Human Rights Watch (2025). "All I Know Is I Want Them Home": Disproportionate Removal of Aboriginal Children from Families in Western Australia. HRW Report.

9.Centre for Excellence in Child and Family Welfare (2025). The impact of homelessness on children, young people and families.

10.Arnautovska, U., Sveticic, J. & De Leo, D. (2014). "What differentiates homeless persons who died by suicide from other suicides in Australia?" Social Psychiatry and Psychiatric Epidemiology, 49(4), 583-9.