Gerontechnology Platform Selection: A Scorecard Approach

A demonstration is not a selection process

Most platform decisions in care homes happen backwards. A vendor demonstrates, the interface looks modern, one manager likes the reporting, and a purchase follows. The decision then has to be defended against a question nobody asked in advance: what problem was this bought to solve, and how would the home know whether it worked?

A defensible process runs the other way. Requirements are written before any demonstration, a pilot measures a short list of things, and the contract answers the questions that only appear twelve months later, when the funded period ends and the data has to come out. What follows is that process in the order it happens.

Write the requirements before you book a demonstration

The material for the requirements is already in the home: incident records, staffing rotas, the complaints log, and the night shift's own account of what interrupts it. A product category is a poor starting point. A scene is a good one.

• The events the platform should shorten: which incidents, in which zones, at which hours, drawn from the home's own records.
• Who will operate it: the staff on duty at 3am, not the project team.
• What it must connect to: the care plan, the medication records, the incident report, and whoever answers an alert.
• What a shift must do when an alert turns out to be wrong, which is a routine state rather than an exception.
• What has to be true for the home to switch it off.

Total cost of ownership does not stop at the funded period

The Innovation and Technology Fund for Application in Elderly and Rehabilitation Care subsidises eligible service units to procure, rent or trial technology products. The Government allocated $1 billion in 2018 to set up the fund and injected a further $1 billion in 2024-25, and the fund's scope now extends to products suitable for household use [I&T Fund]. As at August 2024 the fund had approved grants totalling about $750 million, subsidising about 2,000 service units to procure or rent over 21,000 technology products [LegCo paper, 14 October 2024].

The subsidy covers more than the device. It also covers staff training on the products concerned, and product warranty and maintenance for a maximum of five years [LegCo paper]. Five years is the horizon to plan against: a platform bought with a subsidy has a life after the subsidy, and the licence, connectivity, support and replacement costs land on the home.

The fund's guidance shows a home what a defensible purchase looks like before the home is asked for anything: one product per application; a quotation from a company registered in Hong Kong that lists the price and quantity of each component or service item; and a product that complies with Hong Kong product safety law, with use that complies with the Residential Care Homes (Elderly Persons) Ordinance (Cap. 459), the Private Healthcare Facilities Ordinance (Cap. 633), the Residential Care Homes (Persons with Disabilities) Ordinance (Cap. 613) and the Personal Data (Privacy) Ordinance (Cap. 486) [Guidance Note on the application form]. Duplication of government funds to cover the same project is not allowed. A home that applies the same three tests to a purchase it funds itself will be harder to sell to.

• The purchase price or the rental, per unit and in total, over the funded period.
• Installation, network and power work, which a product quotation often leaves out.
• Training, and the shifts spent inside it.
• Licence, support and replacement after the funded period.
• Staff time per alert, which is the cost the false alarm rate creates.
• Removal and data export at the end, which is rarely free.

A pilot needs exit criteria, and they are written first

A pilot that ends with a general feeling of satisfaction has measured nothing. Before the first unit is installed, decide what number would end the trial early, what number would justify the purchase, and who is allowed to make that call.

One rule makes a pilot fair: the numbers are collected the same way for the new system and for the current arrangement. A platform that beats an unmeasured baseline has proved nothing, and a platform judged on a week of enthusiasm will be judged again, less kindly, a year later.

The baseline is also cheap to collect. Time the current arrangements on the same zones for two weeks, count the handovers where an incident was reconstructed from memory, and count how often a night shift walked to a bed because something looked wrong on a screen or sounded wrong on a monitor.

• Alerts per zone per week, with false alerts counted separately from real ones.
• Time from a real event to a staff member knowing, measured on the same zones as the arrangement it replaces.
• Alerts per shift against the staff on that shift, so workload is a number rather than an impression.
• The share of alerts the system could not classify, which predicts whether staff will trust it.
• Whether the platform's records and the home's existing records agree at the end, without anyone re-typing.
• A named exit date and a named decision owner.

The false alarm rate is an operational cost

A false alarm is cheap once and expensive in aggregate. Each one costs a staff member the walk, the check, the reassurance and the entry in the log, and each one teaches the shift something about that alert which they will remember at the next one.

The effect is behavioural. A 2025 scoping review of 32 publications in BMC Nursing describes alarm fatigue as a form of desensitisation caused by factors including a high number of false alarms, leading to reduced responsiveness and to delayed or inadequate responses to alarms [alarm fatigue scoping review]. That review is drawn from hospital settings, so it should be read as a mechanism rather than as a care home measurement, but it is the mechanism a care home buys into when it installs a platform.

The practical consequence is that the false alarm rate belongs in the cost model rather than in an appendix. Alerts per shift per staff member, and the share of those that were not real, are two numbers worth writing into the pilot report and into the review a year later.

Resident and staff data under the privacy law

A monitoring or care-management platform collects personal data about residents and, in some configurations, about staff. Under the Personal Data (Privacy) Ordinance (Cap. 486), personal data is information relating to a living individual from which it is practicable to identify that individual, and the data protection principles govern how it may be collected, used, retained and secured [PCPD, The Ordinance at a Glance].

Where cameras are part of the proposal, the Privacy Commissioner for Personal Data published a revised Guidance on the Use of CCTV Surveillance in October 2025. It states that a camera system with recording functions that capture images of individuals may involve the collection of personal data; that collection must be for a lawful purpose directly related to the organisation's functions and by fair means; that the data collected must be necessary, adequate but not excessive; that a privacy impact assessment should be carried out before installation; that affected individuals should be told, with conspicuous notices naming the operator and the purpose; that recordings should be deleted as soon as practicable once the purpose is fulfilled; and that footage should be stored in encrypted form, at rest and in transit, with access logged. It also recommends regular reviews of whether the system is still necessary, and states that where the use is no longer necessary, or a less privacy-intrusive alternative would serve the purpose, the use should be discontinued.

The Code of Practice for residential care homes adds a rule written for a home rather than for an office: a home should use a resident's personal data only for the purposes for which it was collected, or for a directly related purpose, and should not display personal data publicly alongside residents' names [Code of Practice 8.10]. A platform that puts a resident's name and alert history on a screen at the end of a corridor has made a design decision with a privacy consequence, and it is worth asking who can see every screen before the contract is signed.

The fund's guidance makes compliance part of eligibility: the use of a funded product has to comply with the Personal Data (Privacy) Ordinance among other legislation [Guidance Note on the application form].

The scorecard

A scorecard is only useful if the evidence column is filled before the vendor arrives. Weight the criteria in the home's own order and write the weights down, even roughly. A home that cannot accept a false alarm on the night shift weights the workload row differently from one that cannot accept a data export problem, and both are right.

Two rows deserve more weight than they usually get. Safety and conformity evidence is the one place where a home can verify hardware claims directly, because the Social Welfare Department's own guidance for gerontechnology products lists the documents to ask for: safety standard test certificates, test reports covering the whole product, and a manufacturer's declaration of conformity showing the model, the standards applied and the certifying body [Considerations and Precautions for Gerontechnology Products]. Data export and exit is the row that decides how much leverage the home still holds in year four.

Questions a vendor should answer in writing

These are worth asking before a pilot rather than after one. A vendor who will answer them in writing is a better risk than one who answers them confidently in a meeting.

• Which standards was this exact model tested against, and will you provide the test reports and the declaration of conformity?
• Where is resident data processed, and does any of it leave Hong Kong?
• What is the default retention period for events and for any footage, and who can change it?
• Can the home export the record in a usable format, on demand, without your software?
• What happens to the home's data at the end of the contract, and what does that cost?
• What is the false alarm rate on a case mix like ours, and how is it measured?
• How many alerts per shift does the system generate at our size, and who is expected to attend them?
• Which of our existing records does it write to, and which does it only read?
• Who trains our staff, for how long, and at whose cost?
• What is the support route outside office hours, and what is the stated response time?

What happens at the end of the funded period

The funded period ends on a date that is known at the start, so build the decisions into the calendar: a review at the end of the pilot, a cost review before the subsidy ends, and an annual check that the platform is still doing what the requirements asked of it.

None of this is unusual procurement discipline. It is the same discipline as buying any system the home will depend on. What is different is who it protects: in a care home the record outlives the contract, and a resident's care plan may have to be reconstructed from it years later.

• Ownership: the care record belongs to the home, in a format it can read without the vendor's software.
• Export: ask for a sample of the export file at procurement, not at termination.
• Replacement: assume the platform will eventually be replaced, and keep the record portable enough for that to be possible.