Time to Care Methodology | Formulas, Privacy and Limits

Purpose and boundary

Carevo is a provider connection platform. It is not a care provider, plan manager or government agency, and it does not see plan budgets, claims, invoices, fund utilisation or care outcomes.

Time to Care measures the demand entering Carevo and the supply response visible on Carevo. It does not estimate total demand in the population. It must not be used as advice, a provider quality measure, an endorsement, or proof that care began.

The present release uses a rolling 180-day window. A rolling window stabilises thin areas while remaining recent enough to detect meaningful access changes.

Component definitions

The four intended components are defined below. Where the current database cannot support the intended meaning, Carevo publishes a clearly named proxy or no value at all.

1. Time to First Contact

Intended formula: median(min(provider_reply_at − enquiry_created_at)) for eligible enquiries receiving a verified provider reply within 30 days.
Unit: Minutes. Lower is better.
Minimum: 30 enquiries with a verified reply in the geography or geography-by-service cell, and at least 30 eligible enquiries overall.
Current release: Not measured. Carevo currently publishes Time to First Provider Connection: median(first non-revoked provider connection at − enquiry created at), capped at 30 days. A connection means a provider chose to receive contact details. It is not a verified reply.
Known limitations: The median among replied enquiries hides enquiries that never receive a reply, so it must be read with the connection rate. Platform notifications, provider subscriptions, service mix, time of day and Carevo workflow changes can affect it.

2. Coverage Gap Ratio

Formula: eligible Carevo enquiries in region R for service S during window T ÷ distinct active listed providers serving R and listing S.
Unit: Carevo enquiries per active listed provider per 180-day window. Lower indicates less platform demand per listing.
Minimum: 30 enquiries and at least 3 distinct active listed providers in the cell.
Provider rule: A provider must have status=active, a non-demo profile, an active service-area record in the geography and, for a service cut, a matching normalised service category.
Known limitations: This is a Carevo directory ratio, not providers per resident or participant. Profile and service-area completeness vary. “Active” and “listed” do not prove capacity, willingness to accept a client, registration, appointment availability or service delivery. The denominator is a release snapshot, not a historical daily inventory.

3. Provider Acceptance Rate

Intended formula: explicit accepted matched-provider opportunities ÷ (explicit accepted + explicit declined opportunities) × 100. One provider-enquiry pair counts once; outcomes must be recorded within 30 days.
Unit: Percentage points. Higher is better.
Minimum: 30 provider-enquiry pairs with an explicit accepted or declined outcome, within a cell containing at least 30 eligible enquiries.
Acceptance definition: The provider confirms they can proceed to the next service step, such as assessment or booking. It does not mean the person chose that provider or care was delivered.
Current release: Not measured. Existing connections and follow-up labels do not consistently record a provider's initial accept-or-decline decision. Carevo does not convert a connection into an “accept”.
Known limitations: Providers may accept conditionally, decline for reasons unrelated to capacity, or leave an outcome unrecorded. Any result will reflect Carevo's matched opportunity mix rather than every request a provider receives.

4. Language-Matched Availability

Intended formula: language-requested enquiries with at least one eligible provider matching language, service and region at enquiry time ÷ all language-requested enquiries in the cell × 100.
Unit: Percentage points. Higher is better.
Minimum: 50 language-requested enquiries in the cell, at least 30 eligible enquiries overall, and a minimum 60% provider-profile language completion rate for the relevant provider pool.
Availability rule: A provider must be active at the event time, list the requested language and service, serve the region, and be eligible to receive the opportunity. This is discoverable platform availability, not guaranteed staff availability.
Current release: Insufficient data. Only 45 mapped enquiries in the window recorded a requested language, and only 0.8% of active provider profiles recorded a language.
Known limitations: Language may be entered in different forms, multilingual staffing changes, proficiency is not assessed, and people may omit a preference. Carevo will never infer language from a person's name, condition or suburb.

Proposed composite score, and why it is withheld

If all four components become publishable, Carevo could place each metric on a 0–100 scale within the same service and release. Lower-is-better measures would use 100 × clamp((P95 − x) ÷ (P95 − P5), 0, 1); higher-is-better measures would use 100 × clamp((x − P5) ÷ (P95 − P5), 0, 1). P5 and P95 would limit the effect of outliers.

Proposed Time to Care composite weights
Component	Weight	Reason
Time to First Contact	35%	Families experience delay directly and the event can be measured precisely once replies are instrumented.
Coverage Gap Ratio	30%	Captures structural pressure between platform demand and listed supply.
Provider Acceptance Rate	25%	Distinguishes a quick initial connection from a provider actually able to proceed.
Language-Matched Availability	10%	Important for equitable access, but applicable to a smaller explicitly requesting subgroup.

Recommendation: do not publish the composite now. It would hide component sample sizes, service mix, missing outcomes and the fact that Carevo users are not population-representative. It could also turn small operational differences into a misleading electorate league table. The score remains withheld until every component meets its quality gate and independent sensitivity testing shows that reasonable weighting changes do not materially reorder regions.

Geography and crosswalk

Normalise the enquiry suburb, state and postcode, with raw form and page-context fields used only as documented fallbacks.
Match that locality to the Carevo suburb collection and use its recorded longitude and latitude as a locality centroid.
Allocate the centroid independently to an ABS ASGS Edition 3 SA3, an ABS Greater Capital City Statistical Area and the AEC federal electorate boundaries applied at the 2025 federal election.
Analyse the internal cell at SA3, then publish only national, capital-city and electorate aggregates. SA3 rows are not included in public downloads.

A suburb can cross an SA3 or electorate boundary. Centroid allocation is therefore a statistical approximation, not address-level geocoding. Direct point-in-polygon allocation is used for each boundary system; Carevo does not pretend an SA3-to-electorate correspondence is exact. Unmatched localities are excluded from geographic results and reported in release diagnostics.

Privacy and suppression

Every published numeric geography or geography-by-service cell requires at least 30 eligible Carevo enquiries.
Time and connection-rate numerators have their own minimum of 30 outcomes. Coverage also requires at least 3 distinct providers.
Language results require at least 50 language-requested enquiries and the provider-language completeness gate described above.
A cell below threshold displays exactly “insufficient data”. It receives no number, range, percentage, ranking, colour scale, rounded proxy or tooltip value.
Suppressed region CSV files contain headers but no hidden values. If future totals could reveal one or two suppressed partition cells by subtraction, at least one additional small published cell must be suppressed.
Carevo never publishes a cut combining condition with a small geography. Condition-by-electorate, condition-by-SA3 and condition-by-locality outputs are prohibited regardless of apparent analytical value.
No row-level enquiry, provider contact, free text, condition, age, name, address or precise coordinate is exported.
The rolling 180-day window stabilises thin samples. Thresholds are re-applied from scratch on every release.

Eligibility and exclusions

The current release includes non-demo Carevo records classified as a service request, aged care enquiry or referral during the rolling window. It excludes known duplicates and senders classified as provider representatives, job seekers or training providers.

Every MongoDB query over enquiries, providers and provider service areas explicitly excludes isDemo=true. Services are normalised through a version-controlled mapping. Changes to eligibility or mappings require a methodology version change and a regenerated release.

Platform-derived bias

These limitations are part of the product, not fine print:

Carevo enquiries are a convenience sample of people who used Carevo. They are not representative of all Australians seeking care.
Provider coverage comes from the Carevo listed directory and active service-area records. It does not prove current capacity, service acceptance, appointment availability, or service delivery.
A provider claim means the provider chose to receive contact details. It does not prove the provider replied, accepted the person as a client, or delivered care.
Suburb and postcode are allocated to SA3, capital-city region and electorate using the recorded locality centroid. Suburbs can cross boundaries, so this is a statistical approximation.
Service categories are normalised from Carevo classifications and provider profiles. Classification and profile completeness vary over time.
Requested-language capture began recently and provider language recording is incomplete. Language results are withheld.
Comparisons can reflect Carevo marketing, search visibility, referral sources, provider notification rules, subscriber behaviour and service mix, not only underlying market access.
Federal electorate results are designed to make local evidence usable by the public and elected representatives. That publication choice does not make the sample representative of every person in the electorate.
Public NDIS, ABS and register data may add context, but it must not be used to “correct” Carevo demand into a population estimate without a validated statistical model.

Public data joins

The first release does not blend public datasets into the component measures. Planned joins are contextual fields with their own source date and geography:

NDIS participant counts and committed supports, used as regional context, not a proxy for Carevo enquiry demand or fund utilisation.
ABS population and SEIFA, labelled by source year and never interpreted as individual disadvantage.
NDIS Quality and Safeguards Commission register and the aged care Provider Register, used to state registration provenance only.

Registration means the relevant regulator records the organisation as registered for specified activities. It is not Carevo vetting, a Carevo endorsement, proof of quality or proof of current availability.

Governance, review and corrections

Method owners: Andre Smith, Co-founder and CEO, and Buddhika Jayawardhana, Co-founder and CTO. Neither is presented as a clinical reviewer.
Refresh: Quarterly public release, with monthly data-quality monitoring. The annual report uses the latest four quarterly releases and documents any break in series.
Change control: Formula, eligibility, service taxonomy, boundary or threshold changes require a new methodology version. Data-only refreshes receive a new dataset ID.
Corrections: Report suspected errors through the contact form with the affected URL and evidence. Carevo investigates reproducible data or code issues, regenerates affected files and records material corrections below.

Families never pay Carevo. Providers pay a subscription to connect with families, with no commission or paid ranking. How Carevo works

Correction log

No material corrections recorded. Methodology first published and reviewed 12 August 2026.