Data and Evidence for Ageing in Kenya: Building Better Intelligence for Long-Term Care Planning

Kenya already knows considerably more about its older population than a simple national headcount suggests. The Kenya Population and Housing Census provides age, sex, household, disability and geographic information. Demographic and health surveys add evidence about health, functional difficulty, insurance and household circumstances. Social protection systems hold administrative information about programme beneficiaries. Health services generate clinical and utilisation data, while Community Health Promoters increasingly create household-level information through digital community-health systems.

Yet a different question arises when Kenya begins planning long-term care: can these sources show who needs continuing support, what families are already providing, where unmet need is concentrated and whether services are helping people remain independent? Within the Kenya Ageing, Long-Term Care & Community Support Knowledge Hub, this distinction is fundamental. Population data can describe ageing without necessarily measuring care need.

The 2019 census counted about 2.74 million people aged 60 and above, representing close to 6% of the population. It also demonstrated substantial county variation: older people accounted for more than 10% of the population in counties including Murang'a, Nyeri and Vihiga, while their population share was considerably lower in several northern counties and Nairobi City. Those differences matter, but age distribution alone cannot tell a county how many people require help with bathing, mobility, meals, medication, communication or supervision.

Kenya's next evidence challenge is therefore not simply collecting more data. It is developing intelligence that connects demography, function, health, disability, family support, social protection and service use strongly enough to guide decisions.

Kenya has an important evidence base, but long-term care asks different questions

The Kenya National Bureau of Statistics sits at the centre of the country's official statistical system. The 2019 census created a particularly important baseline for ageing because its analytical work examined older and vulnerable populations alongside disability, household composition, housing, migration, gender and other social characteristics.

This allows ageing to be understood geographically rather than as a single national trend. Counties differ in population structure, urbanisation, household patterns and access to infrastructure. Those differences shape the type of support likely to be needed.

The Kenya Demographic and Health Survey adds another perspective. Its use of functional domains including seeing, hearing, communicating, cognition, walking and self-care illustrates why disability and functional information can reveal needs that diagnostic information alone misses.

Administrative systems add further layers. The State Department responsible for social protection and senior citizen affairs holds information associated with older-person programmes. The National Council for Persons with Disabilities maintains disability registration and programme information. Health systems record encounters and treatment, while community-health infrastructure creates increasingly detailed household-level data.

Each source has value. None should be mistaken for a complete long-term-care dataset.

A cash-transfer register primarily tells government who is enrolled in a programme. A hospital record describes healthcare activity. Disability registration reflects a defined administrative process. Census data provide population intelligence at intervals. Community-health data are generated for community-health purposes.

Long-term-care planning requires these sources to be interpreted together without pretending that they measure the same thing.

The missing variable is often functional ability

Age is an important planning variable, but it is a weak proxy for individual care need.

Two people aged 78 may have completely different lives. One may farm, travel independently and manage a household. Another may require daily assistance following a stroke. A third may be physically independent but need increasing supervision because of cognitive change.

Long-term-care intelligence therefore needs to move beyond counting older people towards understanding functional ability and support requirements.

Relevant dimensions can include:

• mobility and ability to move safely inside and outside the home;
• ability to manage personal and everyday activities;
• cognition, communication, hearing and vision;
• the availability and sustainability of family or other informal support;
• housing, transport and environmental barriers; and
• changes in function over time rather than a single assessment.

This does not mean creating an intrusive national database containing every detail of older people's lives. Data collection should remain proportionate to a legitimate policy or service purpose.

It does mean recognising that disease, disability, age and care dependency are related but not interchangeable concepts.

Kenya's existing disability evidence demonstrates the importance of this distinction. Functional difficulty becomes more common with age, but not every older person experiencing reduced mobility, hearing, vision or cognition will necessarily appear in a disability register. Conversely, a person registered as having a disability may remain highly independent and require no continuing personal assistance.

For service planning, outcomes-focused support requires information about what people can do, what matters to them and what assistance enables them to maintain everyday life.

A county knows how many older people it has but not how many need support

Consider a county preparing an ageing strategy. Census information shows the number and geographic distribution of residents aged 60 and above. Social protection data show the number receiving relevant cash transfers. Health facilities can describe common conditions and service use.

Officials nevertheless struggle to answer a practical question: how many older residents currently depend on another person for essential daily activities?

If planners use age alone, they risk substantially overstating dependency because many older people remain independent. If they use only formal service utilisation, they may understate need because much support is provided privately by families and never appears in a service record.

The county therefore begins combining existing demographic information with structured evidence from community and health services. Rather than attempting an immediate universal assessment, it identifies indicators that may justify further attention: substantial functional difficulty, repeated falls, recent discharge, cognitive deterioration, living alone without reliable support or severe pressure on an unpaid caregiver.

The result is not a perfect count of long-term-care need. It is a more useful planning picture.

That distinction matters. Good intelligence does not require pretending uncertainty has disappeared. It requires making the uncertainty visible enough that decisions can improve.

Households contain much of the care system that administrative data cannot see

One of the greatest evidence challenges in Kenyan long-term care is that much of the system operates outside formal services.

A daughter helping her mother wash each morning does not generate a care-service transaction. A son sending money to support an older parent may appear in financial flows but not in care data. A neighbour checking on an older person living alone may be crucial to continuity without appearing in any administrative system.

This invisible infrastructure matters because family capacity affects formal demand.

Two older people with comparable functional difficulty may require very different levels of external support depending on whether they live with relatives, whether those relatives are available during the day, whether they can safely provide physical assistance and whether caring responsibilities are sustainable.

Household structure is therefore relevant but cannot simply be equated with support. Living with several relatives does not prove that reliable care is available. Equally, living alone does not necessarily mean somebody is unsupported.

Evidence needs to examine the actual care relationship.

This connects long-term-care intelligence with family partnership and carer support. Understanding who provides care, how much they provide and where arrangements are becoming unsustainable can reveal future demand before a household reaches crisis.

Gender needs to remain visible in ageing data

National averages can conceal important differences between older women and older men.

Women tend to live longer and consequently form a larger share of older populations at advanced ages. Kenyan census analysis has also shown gender differences in disability and household circumstances among older people.

The consequences extend beyond demography.

Women may reach older age after longer periods of unpaid family work and with different lifetime income or asset patterns. Older women may be widowed and living alone. At the same time, daughters, daughters-in-law and other women continue to provide substantial amounts of unpaid support to older relatives.

A long-term-care dataset that counts only service recipients can therefore miss two forms of gender inequality: unequal circumstances among people requiring care and unequal distribution of the work involved in providing it.

Stronger intelligence should allow decision-makers to examine care need and caregiving by sex, age, location, disability and household circumstances where data are available and ethically appropriate.

This is not disaggregation for its own sake. It helps reveal whether apparently neutral policies have different practical effects across populations.

Organisations examining similar evidence questions can use the Social Value Report Builder to structure thinking about outcomes, inclusion and evidence. It is a generic analytical framework rather than a Kenyan statistical or reporting instrument.

County variation should change decisions, not merely appear on maps

Devolution makes geographic intelligence particularly important in Kenya.

National government has responsibilities for national policy and standards in areas relevant to ageing and social protection, while county governments hold important responsibilities across devolved health services and local implementation. Long-term-care development will consequently take place within very different county contexts.

An older population concentrated in densely populated urban neighbourhoods creates different service-design questions from one distributed across remote rural settlements. Travel time, transport, workforce availability, health infrastructure and family migration patterns all affect how support can be organised.

The 2019 census already demonstrates substantial variation in the proportion of older people across counties. But percentage share needs to be interpreted alongside absolute population numbers. Nairobi City may have a younger demographic profile proportionately while still containing a large absolute number of older residents requiring diverse forms of support.

Planning therefore needs several geographic lenses simultaneously:

• how many older people live in an area;
• what proportion of the local population they represent;
• where functional limitations and disability are concentrated;
• how far households are from relevant services;
• where formal and informal care capacity exists; and
• which communities appear underserved relative to estimated need.

Geographic evidence becomes operational when it changes resource deployment.

A county may discover that establishing one central service creates long journeys for most of the population it intends to reach. Another may find that strengthening community and home-based support around several Primary Care Networks offers a more realistic model. Urban areas may require neighbourhood-level analysis because county averages conceal substantial inequality within the same city.

This is where demographic evidence becomes service intelligence.

Health data can reveal pressure but not automatically explain it

Older people often interact with healthcare before any formal long-term-care service exists. Hospital admissions, outpatient attendance, chronic-disease management and community-health encounters can therefore provide important signals.

Repeated healthcare use may indicate unstable health, but it can also reveal problems outside the clinical pathway.

An older person may return to hospital because rehabilitation was unavailable after discharge. Medicines may be clinically appropriate but difficult to manage without household support. A fall may reflect frailty, unsafe housing or unsuitable equipment. A person living with dementia may repeatedly reach emergency care because family caregivers have no practical support when distress escalates.

Health utilisation data become more valuable when analysts can investigate these patterns rather than simply counting episodes.

This requires caution about causation. A high admission rate among older people does not prove inadequate community care. Nor does a lower admission rate automatically indicate successful prevention; barriers to healthcare access could produce the same pattern.

Good intelligence therefore combines quantitative signals with contextual investigation.

The wider principle behind root cause analysis and thematic learning is relevant: recurring patterns should prompt investigation into underlying mechanisms rather than immediate assumptions.

Repeated hospital returns reveal a pathway problem

A county referral hospital notices that some older stroke patients return within weeks of discharge. Initially, the issue appears to be a hospital-readmission problem.

A closer review links discharge information with follow-up evidence from community and primary healthcare services. Several patients have encountered similar difficulties: family members received limited practical preparation, mobility support was inconsistent and access to rehabilitation after returning home varied considerably by location.

The hospital alone cannot resolve all of these issues. Neither can community services if they are unaware that the pattern exists.

The data become useful when they create a shared question: what happens to older people after they leave the facility?

The county begins examining discharge destination, functional status, follow-up, rehabilitation access and subsequent healthcare use together. It does not assume that every readmission is preventable, but it becomes possible to distinguish isolated clinical events from recurring pathway weaknesses.

Over time, the evidence can inform discharge practice, workforce development and local rehabilitation capacity.

This is a stronger use of data than simply publishing a readmission percentage. Intelligence has connected an outcome with the service pathway capable of influencing it.

Social protection data describe programme reach, not the whole population in need

Kenya's social protection infrastructure provides another important source of evidence about older and vulnerable people.

Administrative data can show who is enrolled, payments made and the geographic distribution of programme beneficiaries. Such information is essential for programme management and accountability.

It should not be interpreted as a register of everybody requiring long-term care.

Eligibility for an income-support programme and need for personal assistance are different concepts. An older person may require extensive care without being represented in a particular administrative dataset. Another person may receive social protection while remaining functionally independent.

The same caution applies to disability registration. Kenya's National Council for Persons with Disabilities maintains a formal registration system, but registration data and population prevalence measure different things. The 2022 Kenya Demographic and Health Survey, for example, collected information across functional domains, while census disability analysis used its own population methodology.

Differences between datasets are therefore not necessarily errors. They may reflect different definitions, questions, thresholds and purposes.

Long-term-care planners need metadata as much as numbers: what does each dataset measure, who is included, when was it collected and what conclusions can reasonably be drawn?

Community-level information could provide an earlier view of changing need

Kenya's Community Health Promoter network creates an important opportunity because information can originate close to where people live.

Community Health Promoters are not long-term-care assessors and should not be transformed into a substitute care workforce. Their household contact nevertheless places them in a position where relevant changes may become visible earlier than they would through hospital data alone.

With appropriate training, referral pathways and data governance, community-level systems can contribute to recognising issues such as reduced mobility, missed healthcare follow-up, increasing caregiver strain or an older person becoming unable to manage an established routine.

The value lies in connecting recognition to response.

Collecting information about a problem without creating a route for action can increase administrative workload while changing little for the person concerned.

Future ageing intelligence should therefore distinguish between data needed for national population planning and information required to trigger individual service responses. The first supports policy. The second supports care. Some information may contribute to both, but the purposes should remain explicit.

The Commissioner Evidence Builder can help organisations exploring comparable service-development questions structure evidence requirements, performance expectations and assurance. In a Kenyan context it should be treated as a generic planning tool, not as a description of county purchasing arrangements or an official national framework.

Interoperability matters because the same person appears in several systems

An older Kenyan may simultaneously appear in health, social protection, disability and other administrative systems. The danger is assuming that the answer is one universal database.

It is not.

Different systems collect information for different lawful purposes, and access should remain controlled. The stronger objective is appropriate interoperability: enabling relevant information or aggregated intelligence to connect where there is a legitimate reason while preserving privacy and security.

Kenya's Digital Health Act 2023 and developing national digital-health architecture make this particularly important. Health information is becoming increasingly structured and connected, but long-term-care intelligence extends beyond health.

The policy challenge is deciding which relationships between datasets create genuine public or individual benefit.

For example, aggregated analysis might examine whether older people in particular areas experience high hospital use alongside weak rehabilitation availability. Service-level information might allow a legitimate referral to move between parts of a pathway. National analysts may need population projections without accessing identifiable individual records.

These are different data uses and require different governance.

The broader discipline of interoperability and system integration is therefore as much about governance and standards as technology.

Privacy and trust are part of the evidence infrastructure

Long-term-care data can reveal intimate aspects of a person's life: ability to wash or use the toilet, cognitive changes, family relationships, financial vulnerability, disability, home conditions and safeguarding concerns.

Kenya's Data Protection Act 2019 provides the wider legal framework for personal-data protection, while sector-specific requirements apply to digital health information. Future long-term-care information systems need to develop within those protections.

More data are not automatically better data.

Collecting information because it might someday be useful can undermine proportionality and public trust. Sharing identifiable information between agencies simply because technology permits it can create further risk.

Strong governance asks why information is required, who needs access, how accuracy can be challenged, how long information is retained and what happens when a breach occurs.

Older people also need accessible explanations. Consent processes that exist only as lengthy digital notices provide limited protection where people cannot realistically understand them.

Trust has operational value. People who believe information may be misused may withhold it, producing poorer data and potentially poorer care.

Kenya needs longitudinal intelligence, not only snapshots

Long-term care is fundamentally about change over time.

A census provides an essential population baseline, but it is periodic. A single assessment describes one point in a person's life. Service-use data show encounters. Long-term-care planning needs to understand trajectories.

How quickly is the older population growing in different counties? At what ages does functional difficulty become more prevalent? How does migration alter the availability of family support? Which health events most commonly precede substantial dependency? How long do people require support, and how often do they regain function after illness or injury?

These questions require longitudinal evidence, repeated surveys, administrative analysis or carefully designed linkage rather than one-off counts.

Kenya's population projections can support forward planning, but projecting the number of older people is not the same as projecting care demand. Future need will also be shaped by disease patterns, disability, prevention, rehabilitation, housing, family structure, technology and expectations about formal support.

Scenario modelling can therefore be useful where uncertainty is made explicit.

Organisations exploring future demand can use the Digital Twin Scenario Modeller to test how different assumptions about demand, capacity and workforce might affect services. Such modelling does not predict Kenya's future or replace official population projections; its value lies in making assumptions visible enough to test.

A county discovers why projections need more than population growth

A county estimates future home-support demand by applying a fixed percentage to projected growth in its population aged 60 and above. The calculation is straightforward and produces a clear number.

Further analysis reveals why the apparent precision is misleading.

Some sub-counties have substantial out-migration among younger adults, potentially reducing family support. Other areas have stronger multigenerational households. Access to rehabilitation differs geographically. The proportion of people reaching advanced older age is changing, and women form a larger share of some older age groups. Transport costs make identical home-support models much more expensive in dispersed communities.

Instead of abandoning forecasting, planners create several scenarios. One assumes family-care availability remains relatively stable. Another tests lower availability as migration continues. A third models stronger prevention and rehabilitation, while a fourth examines greater formal service uptake as public expectations change.

No scenario is presented as a prediction.

The exercise nevertheless improves decisions because leaders can identify which investments remain useful under several plausible futures: better functional data, stronger community pathways, workforce development and clearer evidence about unpaid caregiving.

Forecasting has therefore become a governance tool rather than a claim to know exactly what will happen.

Quality data should show whether support improves life

As formal long-term-care services develop, Kenya will need evidence not only about demand but about quality.

The easiest measures will be activity: people served, visits completed, beds available, staff employed or money spent. Those indicators are useful for operational management, but they cannot establish whether care is good.

Outcome intelligence asks harder questions.

Does an older person maintain mobility? Can somebody continue participating in family and community life? Has a caregiver's unsustainable burden reduced? Are avoidable falls decreasing? Is somebody able to remain at home because support is reliable? Does a person living with dementia experience greater continuity and less distress?

Different services will require different measures, and not every outcome can be attributed to one provider. Nevertheless, Kenya can avoid building a long-term-care evidence system dominated by inputs and activity from the outset.

This is where quality data and performance metrics need to remain connected to people's lives.

Quantitative evidence should also be complemented by the experience of older people and families. Complaints, feedback, community participation and qualitative research can reveal problems that aggregate performance data miss.

A service can meet its numerical targets while people experience poor continuity, limited choice or inaccessible communication.

Data gaps should be treated as governance information

One of the most useful findings in an emerging long-term-care system may be that an important question cannot yet be answered.

Leaders sometimes respond to missing data by demanding immediate new reporting. That can create large administrative burdens without resolving the underlying evidence problem.

A stronger approach is to maintain a visible evidence-gap framework.

For each strategic question, decision-makers can identify what is known, the source, its limitations, what remains unknown and whether closing the gap would materially improve a decision.

Some gaps may justify new national survey questions. Others may require county-level research, improved administrative coding, qualitative work or better analysis of data already collected. Some will not justify additional collection at all.

The principle resembles quality monitoring systems: evidence should create a route from observation to decision and improvement.

The Governance Maturity Assessment can help organisations consider whether evidence, accountability, escalation and decision-making are sufficiently connected. It is not a Kenyan governmental assessment and should be adapted only as a generic governance aid.

Older people should influence what the system chooses to measure

Data systems can become technically sophisticated while measuring priorities that matter primarily to institutions.

Older people and family caregivers provide a different perspective.

A ministry may want prevalence and expenditure. A county may need geographic demand and workforce information. A provider may monitor service delivery. An older person may care most about whether the same worker arrives, whether she can attend church, whether she can still cook part of her own meal or whether support allows her daughter to remain employed.

These perspectives are not competing. A mature evidence system connects them.

Meaningful co-production and lived experience can help determine which outcomes deserve attention, how questions are asked and whether official measures reflect everyday reality.

This is especially important where data concern disability, cognition, dependency or family care. Poorly framed measures can portray older people only through deficits and burdens. Better measures can capture capability, participation, relationships and the support that enables independence.

Community evidence changes the interpretation of a performance measure

A new county support programme reports strong performance because a high proportion of planned household contacts are completed each month.

Older people participating in a community review describe a more complicated picture. Some value the contact, but others explain that visits are brief and repeatedly cover the same questions. Families say that concerns are recorded but they rarely know what happens next.

The county does not discard the activity measure. It changes its interpretation.

Completed contact remains useful for monitoring reach, but additional evidence is introduced around referral completion, changes in identified need, continuity and the person's experience of support.

Managers also examine cases where the same issue is repeatedly recorded without resolution.

The result is a more credible evidence framework because performance is no longer defined simply by what the service does. It also considers what happens afterwards.

This illustrates an important principle for Kenya's emerging long-term-care system: the earlier outcomes are built into information architecture, the less difficult it will be to retrofit them after activity measures have become institutionalised.

National and county intelligence need a shared language without eliminating local variation

Kenya does not need every county to provide identical long-term-care services. Different populations and geographies require different operational responses.

But comparison becomes difficult if fundamental concepts mean different things everywhere.

National development of long-term-care intelligence could therefore establish a small common language around concepts such as older person, functional need, informal caregiving, service type, outcome and safeguarding while allowing counties to collect additional information relevant to local priorities.

Common definitions make aggregation and comparison more credible. They also allow persistent variation to become visible.

If one county appears to have much lower need than neighbouring areas, decision-makers can investigate whether the difference is genuine or reflects access, recording or definitions. If one pathway produces better outcomes, analysts can explore why rather than assuming the model can simply be copied.

This creates a learning system rather than a reporting hierarchy.

The proposed Older Persons Bill published in 2024 illustrated the direction of policy thinking by including provisions concerning national data, research and information alongside county registries. As a proposed legislative measure, it should not be treated as an already implemented national long-term-care data system. Its significance lies in recognising that information infrastructure is part of building an older-person support system, not an administrative afterthought.

Future statistical reform creates another opportunity

Kenya's statistical framework is itself evolving. The Statistics Bill 2026 proposes a replacement framework for official statistics, reflecting constitutional devolution, data protection, new data technologies and changing forms of evidence. At the time of writing, it remains proposed legislation rather than an enacted replacement for the existing Statistics Act.

For ageing policy, the wider direction is important.

Future evidence will increasingly combine traditional official statistics with administrative data, digital systems and potentially other carefully governed sources. Long-term-care intelligence needs to benefit from that expansion without weakening statistical quality or privacy.

The challenge will be to preserve clear definitions, transparent methodology and appropriate quality controls as information becomes faster and more granular.

Administrative data can be timely but are shaped by service eligibility and access. Survey data can capture people outside services but depend on sampling and question design. Census data offer extraordinary population coverage but cannot answer every specialised care question. Qualitative evidence provides depth but is not a population estimate.

The strongest intelligence system understands what each form of evidence can and cannot establish.

The international lesson is to build evidence before long-term care becomes institutionally fragmented

Countries with established long-term-care systems often struggle to integrate evidence retrospectively because healthcare, social services, insurance, municipalities and providers have developed separate information architectures over decades.

Kenya's institutional context is different. A comprehensive formal long-term-care system is still developing while digital health, community health, disability reform, social protection and national statistical infrastructure are evolving simultaneously.

This creates an opportunity to establish common evidence principles earlier.

The transferable lesson for other countries is not that Kenya should create a single national care database or replicate an international measurement framework. It is that system design and information design should develop together.

If future funding mechanisms, service standards and provider arrangements are created without deciding what outcomes need to be understood, activity reporting can become embedded before meaningful evidence exists. If data requirements become excessively complex, however, frontline workers may spend more time recording care than delivering it.

The stronger balance is a proportionate evidence architecture: enough common information to understand need, equity, quality and outcomes, combined with local flexibility and strong protection of personal information.

Conclusion

Kenya does not begin the development of long-term-care intelligence from an empty evidence base. The census, demographic and health surveys, disability data, social protection systems, health information and expanding community-level digital infrastructure already provide important pieces of the picture. The strategic challenge is that those pieces answer different questions.

A stronger long-term-care evidence system will need to distinguish age from dependency, disability registration from functional need, programme enrolment from population need and service activity from human outcomes. It will also need to make unpaid family care, geographic inequality and changing function more visible without turning older people's lives into an unnecessarily intrusive data exercise.

The most useful direction is therefore connection rather than indiscriminate collection. National statistics can establish population trends. County intelligence can translate them into local planning. Health and community data can identify patterns and changing need. Service evidence can show what support is delivered. Older people and families can help determine whether that support actually improves life.

Governance is what turns those sources into intelligence. Definitions need to be clear, uncertainty acknowledged, privacy protected and evidence connected to decisions about funding, workforce, prevention and service design.

As Kenya's ageing and long-term-care system develops, building that evidence architecture early offers a significant advantage. Better data will not determine every policy choice, but it can make need more visible, assumptions more testable and accountability considerably stronger.