Measuring Quality in Hong Kong Care Services: Moving Beyond Activity, Occupancy and Service Volume

A Hong Kong residential care home can be full, a community service can exceed its annual activity target and a disability programme can deliver every session specified in its service agreement. All three may still need to ask whether people's lives are actually improving.

This is the central challenge of quality measurement. Activity, capacity and utilisation matter because publicly funded and regulated services need to demonstrate that resources are reaching the people for whom they are intended. Yet volume cannot answer questions about independence, continuity, dignity, safety, participation or whether support remains appropriate as people's needs change. Across the wider Hong Kong Ageing, Long-Term Care & Community Support Knowledge Hub, this distinction between service provision and human outcome is fundamental to understanding how an increasingly complex care system should be governed.

Hong Kong already possesses an important measurement architecture. The Social Welfare Department's Service Performance Monitoring System for subvented welfare services distinguishes Essential Service Requirements, Service Output Standards and Service Outcome Standards, alongside the 16 Service Quality Standards. Residential care is also subject to statutory licensing requirements, while individual service models generate their own clinical, social, operational and financial evidence.

The stronger opportunity is therefore not simply to collect more data. It is to ask better questions of the data already being produced: what does each measure tell leaders, what does it fail to show, and what combination of evidence gives the clearest view of quality?

Hong Kong already distinguishes service activity from service outcomes

The architecture of the Service Performance Monitoring System is important because it does not treat all performance measures as equivalent.

For subvented welfare services, Funding and Service Agreements establish service requirements and performance expectations. Essential Service Requirements describe key characteristics or conditions of the service. Service Output Standards measure defined volumes of activity. Service Outcome Standards measure specified effects associated with service delivery.

This distinction creates an important foundation for quality measurement.

An output might tell an organisation how many people attended a programme, how many cases were handled or whether a required level of service was delivered. An outcome asks a different question: what changed for the person, family or service population as a result?

Neither type of measure is redundant. A service that achieves excellent outcomes for a very small proportion of the people it is funded to support may still have a serious access or capacity problem. Equally, a service that reaches large numbers of people but produces little meaningful benefit cannot demonstrate quality through volume alone.

The analytical task is to understand both dimensions simultaneously.

Occupancy is valuable information, but it is not a quality score

Residential care illustrates the problem particularly clearly.

Occupancy matters operationally. It affects resource utilisation, financial viability and the extent to which available residential capacity is being used. In subsidised services, capacity also matters because Hong Kong faces sustained demand for long-term care as its population ages.

Yet high occupancy can coexist with very different experiences of care.

Two homes may each operate close to capacity. One may have stable staff, responsive care planning, strong family communication and residents participating meaningfully in everyday decisions. The other may experience high turnover, rigid routines and repeated delays in responding to changing health needs.

The occupancy figure cannot distinguish them.

Nor should lower occupancy automatically be interpreted as poor performance without context. Temporary vacancies may reflect refurbishment, admission timing, changing service specifications or deliberate restrictions needed to maintain safe provision.

Quality measurement therefore requires leaders to interpret utilisation rather than merely display it.

The same principle applies to day services, home-based support and community programmes. Numbers reveal reach and capacity. They do not independently reveal the quality of what happened within that capacity.

The person should remain the unit of meaning

Organisational measurement systems naturally aggregate information. This is necessary for management, funding accountability and system planning. The risk is that people disappear inside averages.

A service may report 90 per cent satisfaction. That is useful, but it tells leaders little about the remaining 10 per cent unless the underlying experience is explored. It may also obscure people who could not participate in the survey, people with communication difficulties or those who felt uncomfortable criticising a service on which they depend.

Similarly, an average improvement in an outcome measure may conceal a subgroup whose outcomes are deteriorating.

Good quality data and performance metrics therefore need disaggregation where this is proportionate and meaningful. Leaders may need to understand differences by service unit, support need, duration of service or other relevant characteristics rather than relying exclusively on organisational averages.

The purpose is not to generate increasingly complex statistical reports. It is to preserve the link between measurement and the lives the data represent.

Operational scenario: full occupancy hides a change in resident experience

A residential care home for elderly people has operated at very high occupancy for several quarters. There are no major licensing concerns, staffing numbers meet relevant requirements and the home's overall satisfaction results remain positive.

Management therefore regards the service as stable.

A more detailed quality review identifies several changes. Residents are participating less frequently in organised and informal activities. Family comments increasingly mention difficulty speaking with the same staff member twice. The number of residents requiring substantial assistance with mobility has also increased, while the workforce establishment has not materially changed.

No single indicator shows a major failure. The combined pattern suggests that the resident population has become more dependent while the operating model has remained largely unchanged.

The home begins measuring more than occupancy. It examines the proportion of residents participating in preferred activities, continuity of key staff relationships, response times to changes in care needs and the amount of direct staff time required across different parts of the day.

This provides management with a better basis for considering skill mix and deployment.

Occupancy remains on the dashboard because it is still an important operational measure. It is no longer treated as evidence that the service is functioning optimally.

Outcome measures need to describe changes that services can reasonably influence

Outcome measurement becomes weak when organisations choose measures that are either too narrow or too distant from the service's actual contribution.

For example, a short-term rehabilitation or community support service may reasonably examine whether a person regains particular functional abilities, develops confidence, remains at home or achieves agreed goals. It cannot reasonably claim sole responsibility for every subsequent health or social outcome.

Conversely, measures such as attendance or number of contacts are usually too close to the service activity to demonstrate meaningful impact.

A useful outcome measure should therefore have a plausible relationship with the service.

Depending on the service, this might involve:

• maintaining or improving functional ability;
• achievement of individual goals;
• participation in community or everyday activity;
• stability of health or support arrangements;
• carer confidence or sustainable caring arrangements;
• avoidance of preventable deterioration; or
• experience of choice, dignity and control.

These examples should not become a universal Hong Kong outcome set. Different services exist for different purposes, and measurement should reflect those purposes.

The stronger principle is that the measure should describe something that matters to the person and that the service can credibly influence.

Maintenance can be a successful outcome

Care services frequently support people whose underlying conditions are progressive, fluctuating or unlikely to improve substantially. This creates an important measurement problem.

A simplistic outcome framework may reward visible improvement while undervaluing stability.

For an older person with increasing frailty, maintaining mobility for another six months may represent a substantial achievement. For someone with advanced disability, preventing avoidable deterioration and sustaining participation in familiar routines may be a meaningful outcome. For a family carer, continuing to care safely without reaching exhaustion may be more important than an abstract improvement score.

Quality measurement therefore needs to distinguish improvement, maintenance and managed deterioration.

The baseline matters. Without understanding where the person started, a later result may be misleading.

This is one reason outcomes-focused support should begin with clear individual objectives rather than applying identical expectations to everyone using a service.

Quality measurement should separate attribution from contribution

Many outcomes in Hong Kong care services result from several actors working together.

An older person's ability to remain at home may be influenced by family support, a District Health Centre, Hospital Authority services, community care, housing conditions and the availability of a foreign domestic helper. A person with a disability may achieve greater community participation because transport, rehabilitation, employment support, family involvement and social welfare services combine effectively.

No individual provider can reasonably claim full attribution for the result.

This does not make outcome measurement impossible. It changes the language of accountability.

Organisations should be able to explain their contribution: what support they provided, why it was expected to help, what changed and what other factors affected the result.

This is particularly important for integrated and community services, where excessive pressure to attribute outcomes to one organisation can encourage artificial claims rather than meaningful learning.

Individual outcomes and population-level performance serve different purposes

Care organisations need evidence at more than one level.

Individual plans and reviews answer whether support is helping a particular person. Service-level measurement examines whether outcomes are consistent across a programme. Organisation-level analysis asks whether performance varies between services and whether common risks need strategic attention.

These levels should connect without becoming identical.

A detailed individual outcome measure may be valuable for care planning but impossible to aggregate meaningfully. Conversely, an organisation-wide indicator may support governance while offering little value to the frontline practitioner reviewing one person's support.

The measurement system therefore needs hierarchy rather than uniformity.

At the individual level, evidence should help staff and the person understand progress and adjust support. At service level, managers need trends and variation. At organisational level, senior leaders and governing bodies require a smaller set of indicators capable of revealing material risk, quality and change.

Organisations can use the Quality Dashboard Builder as a general method for structuring such an assurance hierarchy. It is not a Hong Kong regulatory framework, and the measures selected still need to reflect applicable Social Welfare Department requirements and the organisation's actual services.

Operational scenario: a community service meets every output but outcomes plateau

A community support programme has consistently achieved its agreed service volume. Referral numbers are healthy, staff deliver the required interventions and participation remains strong.

The organisation also records individual goals. Initial analysis shows that the proportion of people achieving those goals has stopped improving even though activity has increased.

Management first considers whether service quality has declined. A closer review identifies a different explanation. The profile of people entering the programme has changed. More participants now have complex health, mobility and family circumstances, and a greater proportion require longer periods of support.

The existing outcome measure gives the same weight to every participant regardless of baseline complexity or whether maintaining ability is itself a positive result.

The service does not simply replace the measure because its results have become less favourable. Instead, it refines the analysis. Outcomes are reviewed against baseline need, goal type and duration of intervention. Maintenance is recorded separately from improvement, while unresolved goals are examined for common causes.

The revised approach reveals that the programme is still delivering substantial benefit, but that its service population has changed.

This insight has operational consequences. Workforce planning, programme design and expectations about duration of support are reconsidered. The organisation can explain the changing results rather than either hiding them or wrongly concluding that the service has deteriorated.

Service-user satisfaction is useful but should not become a proxy for quality

Satisfaction measures are familiar, understandable and relevant. They can provide valuable evidence about people's experience of welfare services and form part of formal outcome arrangements in some services.

They also have limitations.

People may be satisfied with caring staff while receiving support that does not maximise their independence. Families may value risk reduction while the person receiving support wants greater freedom. Long-term residents may compare their care only with what they have previously experienced rather than with a theoretical alternative.

High satisfaction may therefore coexist with limited choice, inconsistent continuity or low expectations.

Conversely, dissatisfaction does not automatically prove poor-quality practice. A service may need to place reasonable limits around unsafe requests, or a family may disagree with a person's own preferences.

Strong service-user feedback and co-production therefore use satisfaction as one evidence source among several.

The more informative questions often examine specific experiences: whether people understand their support, participate in decisions, know how to raise concerns, experience consistent relationships and feel that staff listen to what matters to them.

Feedback needs to include people who are easiest to miss

Measurement systems can reproduce inequality if they hear mainly from people who communicate easily and engage readily with formal processes.

Some people using Hong Kong disability, dementia and long-term care services may have cognitive impairment, communication differences, sensory impairment or limited confidence in completing written surveys. Others may rely on family members or staff to facilitate participation.

Exclusion from measurement creates a double disadvantage: their experience is less visible, and service design may then respond primarily to people whose views are easier to collect.

Accessible information, observation, supported communication, advocacy and individual review can therefore form part of outcome evidence.

Family views remain important but should not automatically substitute for the person's own perspective where the person can express preferences directly or with support.

Measurement design is consequently a rights issue as well as a technical one.

Workforce measures help explain quality rather than merely describe employment

Quality outcomes rarely change independently of workforce conditions.

Turnover, vacancies, absence, supervision, competence and continuity affect relationships and service delivery. These measures therefore belong in quality analysis even though they are not direct measures of service-user outcomes.

The distinction matters.

A rising turnover rate does not prove that people are receiving poor support. It signals a condition that may affect quality. If turnover rises at the same time as complaints, overdue reviews and weaker continuity, the combined evidence becomes more significant.

Similarly, strong retention should not automatically be treated as proof of competence. A stable workforce may still require development as service needs become more complex.

This is where workforce assurance connects with service measurement.

Organisations with multiple services can also examine whether differences in workforce stability correspond with differences in outcomes. Such analysis can help leaders move beyond assumptions and identify where management practice, specialist skills or deployment may be influencing service quality.

The Predictive Workforce Risk Module offers a general structure for examining relationships between turnover, vacancy, retention and continuity risks. In Hong Kong it should be adapted as an analytical aid rather than treated as a substitute for local employment requirements or service standards.

Safety indicators require denominators and context

Incident numbers are frequently used in quality reporting because they are concrete and measurable. Raw counts can nevertheless be highly misleading.

Twenty falls in a large residential service are not analytically equivalent to twenty falls in a much smaller service. A rise in recorded safeguarding concerns may reflect deteriorating practice, but it can also follow stronger staff awareness and reporting.

The denominator and context matter.

Rates per resident, service user, occupied place or relevant unit of activity may provide a clearer picture than counts alone. Severity should also be distinguished from frequency. Ten minor events and one catastrophic event should not be treated as interchangeable simply because both can be represented numerically.

Trend analysis becomes more valuable where organisations connect incidents with contributing factors and subsequent action.

This is the measurement counterpart to learning from incidents: the objective is not to minimise the number displayed on a report, but to understand whether harm is being prevented and whether known risks are being controlled more effectively.

Operational scenario: a rise in safeguarding reports is initially treated as deterioration

A disability service records a marked increase in safeguarding-related concerns over two reporting periods. At first glance, the trend appears negative and senior management asks whether practice has deteriorated.

Review of the underlying cases shows that most of the increase involves earlier identification of relatively low-level concerns. Staff training delivered several months earlier has improved awareness of financial exploitation, coercion and neglect. Employees are escalating issues that may previously have gone unrecorded.

The organisation therefore avoids setting an arbitrary target to reduce reporting.

Instead, it examines whether concerns are identified promptly, whether appropriate protective action follows, whether the person is involved in decisions and whether repeated themes emerge. It also monitors severe incidents separately.

Over time, the number of initial concerns remains higher than before the training, while recurrence of several previously common problems falls.

The measurement interpretation has changed. More reports are not automatically read as less safety. What matters is the relationship between reporting culture, severity, response and subsequent risk.

This distinction is particularly important in care environments where an apparently low incident rate can mean either excellent prevention or weak recognition and reporting.

Measurement should support improvement rather than encourage gaming

Every performance system influences behaviour.

If organisations are judged predominantly on volume, they may prioritise throughput. If they are judged solely on short-term improvement, they may become reluctant to support people whose needs are complex. If low incident numbers are treated as inherently positive, under-reporting can become an unintended incentive.

No performance framework can eliminate these risks entirely.

The answer is balanced measurement. Indicators should be interpreted together, and leadership should remain alert to unintended consequences.

Good quality monitoring systems therefore contain both measures and challenge. They create an opportunity to ask whether the apparent result reflects genuine improvement or simply a change in recording, case mix or operational behaviour.

Longitudinal measurement is often more valuable than a single snapshot

A quality measure becomes more informative when it shows direction.

One quarterly result may be affected by unusual events, a small sample or changes in the people using the service. A sustained trend provides stronger evidence of whether performance is genuinely improving or deteriorating.

This is particularly important where interventions take time to influence outcomes.

Training may not immediately change service-user experience. A new supervision model may initially increase the number of issues identified before reducing recurrent problems. Digital records may temporarily increase administrative burden while staff learn the system.

Governance therefore needs enough historical information to distinguish temporary disruption from persistent variation.

Longitudinal measurement can also reveal whether an improvement action lasts. A service that corrects overdue reviews immediately after an audit but returns to the previous position six months later has not achieved sustainable improvement.

Measurement should detect recurrence, not simply closure of an action.

Data quality determines the credibility of the conclusion

Increasing the sophistication of analysis does not compensate for unreliable underlying data.

If different service units define an incident differently, comparison becomes weak. If outcome reviews are completed inconsistently, aggregated results may mislead. If missing information is silently excluded from percentages, apparently strong performance can reflect incomplete recording.

Organisations therefore need clear definitions, proportionate validation and transparent treatment of missing data.

Digital systems can help by standardising fields, applying validation rules and reducing manual aggregation. They can also accelerate the spread of poor information if staff enter data inconsistently or system design encourages superficial completion.

This makes digital records, data and information governance part of quality measurement rather than merely an IT matter.

Leadership should be able to ask where an indicator came from, how consistently it is recorded and what limitations should accompany its interpretation.

Operational scenario: a quality dashboard identifies an apparent service outlier

An NGO introduces an organisation-wide dashboard and discovers that one elderly service has a substantially higher rate of falls than comparable units.

The initial response is concern about the quality of care.

Before initiating a major improvement programme, the organisation validates the data. It finds that the service records every assisted descent and near-fall as a fall, while several other units record only events where a resident reaches the floor.

The services are therefore not comparable.

The organisation establishes a common definition and retrospectively reviews a sample of records. Once the data are standardised, the apparent performance gap narrows substantially, although one genuine area of elevated risk remains among residents with particular mobility needs.

Management can now target that problem rather than incorrectly labelling the whole service as poor performing.

The exercise also changes governance practice. Future dashboards identify key indicator definitions, missing-data rates and significant changes in recording methodology.

This example demonstrates a fundamental principle: sophisticated presentation cannot make inconsistent evidence comparable. Measurement quality depends first on the reliability of the data-generating process.

Governing bodies need fewer measures and better explanations

Frontline services may legitimately monitor dozens of indicators. Governing bodies cannot interpret every operational measure with equal depth.

The organisation therefore needs to decide which information signals strategic quality.

A board-level or governing-body view might include service outcomes, regulatory or statutory concerns, material performance variation, significant safeguarding and incident themes, workforce instability, complaints, major audit findings and progress against improvement priorities.

The objective is not to create the shortest possible dashboard. It is to distinguish strategic assurance from operational management.

Where a measure deteriorates, governing-body members need context: why has it changed, who is affected, what action is underway and when should improvement become visible?

The Governance Maturity Assessment can help organisations examine whether information flows and oversight arrangements support appropriate challenge. It does not determine Hong Kong governance compliance but can help structure the underlying assurance questions.

Comparability is useful, but not all services should look the same

Large NGOs may want to compare services across their organisation. Government and system planners also need comparable information to understand capacity and performance.

Standardisation supports this purpose, but excessive standardisation can distort differences between service models.

A residential home, a day service, a rehabilitation programme and a community support service should not necessarily produce identical outcome profiles. Their purposes, populations and durations of support differ.

The strongest measurement frameworks therefore combine a core set of comparable measures with service-specific evidence.

Common organisational measures may include experience, workforce stability, complaints, safety and completion of review processes. Service-specific measures can then examine the outcomes most closely connected with each model.

This approach allows leadership to compare what should be comparable without pretending that every service is interchangeable.

Quality measurement should lead back to resource and service-design decisions

Measurement has limited value if it remains within a quality department.

The findings should influence workforce planning, training, technology investment, service redesign and the use of financial resources.

If outcome data consistently show that a particular group gains less benefit, leaders need to ask whether the service model is sufficiently accessible. If continuity deteriorates as vacancies rise, recruitment and retention become quality priorities. If staff spend excessive time duplicating records, digital or workflow redesign may release capacity for direct support.

This is where quality evidence becomes management intelligence.

Organisations examining how claims about performance are supported can use the Commissioner Evidence Builder as a general framework for connecting assertions, measures and supporting evidence. Within Hong Kong, applicable Funding and Service Agreements, licensing requirements and Social Welfare Department arrangements remain the authoritative service-specific requirements.

Predictive measurement is an emerging opportunity, not a substitute for judgement

As Hong Kong care organisations digitise more operational information, they will increasingly be able to examine relationships across data sets.

A combination of rising absence, reduced supervision, delayed reviews and increased complaints may provide earlier warning of service instability than any single threshold. Analytics can help identify such patterns.

More sophisticated modelling may eventually support forecasting of workforce pressure, demand, deterioration or service risk. Artificial intelligence may also help analyse unstructured information such as comments or recurring themes.

These possibilities should be treated as emerging capabilities rather than established evidence of better quality.

Predictive tools depend on reliable historical data, appropriate interpretation and governance around bias, privacy and false positives. They should support professional judgement rather than replace it.

The stronger immediate opportunity for many organisations is simpler: integrate existing data, establish clear definitions and routinely examine trends rather than waiting for an indicator to breach a formal requirement.

The international lesson is to measure what matters without abandoning accountability

Hong Kong's arrangements are shaped by its own welfare system, subvention model, residential regulation and NGO sector. Other countries cannot transfer the Service Performance Monitoring System directly into different funding and institutional arrangements.

The underlying measurement challenge is widely shared.

Governments need evidence that funded capacity is being delivered. Providers need operational information. People using services need support that makes a meaningful difference to their lives. Regulators and governing bodies need assurance that risk is controlled.

The difficulty arises when one type of measure is expected to answer all of these questions.

Activity measures remain essential but cannot replace outcomes. Outcome measures need context and cannot eliminate qualitative evidence. Satisfaction is valuable but cannot represent every aspect of quality. Workforce and safety indicators explain conditions and risk but should not automatically be treated as direct measures of human experience.

The transferable lesson lies in building a balanced evidence architecture rather than searching for a single perfect quality score.

Conclusion

Hong Kong does not need to choose between operational accountability and meaningful outcome measurement. Its existing care architecture already recognises the distinction between essential service requirements, outputs, outcomes and broader service-quality standards. The opportunity is to connect those forms of evidence more intelligently.

Activity, occupancy and service volume remain necessary. They show whether capacity exists, whether funded services are being used and whether organisations are delivering what was agreed. They become problematic only when they are treated as sufficient evidence of quality.

A stronger measurement model places people's outcomes and experience alongside service volume, workforce stability, safety information, complaints, quality assurance and longitudinal trends. It also recognises maintenance as a legitimate outcome, distinguishes contribution from attribution and validates the data before drawing conclusions.

For governing bodies and senior leaders, the objective should be neither maximum measurement nor a simplistic headline score. It should be enough reliable evidence to understand where quality is improving, where variation persists, what is driving that variation and what operational decision should follow.

As Hong Kong responds to ageing, complex support needs and the growing use of digital systems, that capacity will become increasingly valuable. The strongest care organisations will be those able to demonstrate not only how much support they provide, but what that support changes, protects or sustains in people's lives.