Short Physical Performance Battery

Short Physical Performance Battery in Ageing Trials and Economic Evaluation

The Short Physical Performance Battery, known as the SPPB, times an older person standing in three balance positions, walking a short course at usual pace and rising five times from a chair, and turns each result into points on a 0 to 4 scale. Health economists meet the SPPB as the primary or secondary outcome of trials of exercise, rehabilitation and frailty programmes, and sometimes as the scale that defines the health states of a decision model. The battery carries no preference weights, so a gain of one point says nothing by itself about QALYs or costs, and every economic use rests on a choice about how points become utilities, events or resource use. Guralnik and colleagues showed in 1995 that people over 70 with the lowest scores were 4.2 to 4.9 times as likely to be disabled four years later as those with the highest, which is why a short test done in a clinic corridor can carry weight in a cost-utility analysis.

Three timed tasks and a summary score

The SPPB is a clinical outcome assessment of the performance type, meaning the patient carries out a standardised task and an assessor times it. The National Institute on Aging (NIA), which developed the battery, describes the SPPB as an objective tool for evaluating lower extremity functioning in older persons. The NIA makes the battery available without permission or royalty fees, with training materials that include administration instructions, safety tips and a scoring sheet.

Guralnik and colleagues introduced the battery in 1994 after testing more than 5,000 people aged 71 and over in three communities. Their tests covered standing balance in three foot positions, the time to walk 8 feet and the time to rise from a chair and sit down again five times. Later protocols kept the structure but changed some details. The Lifestyle Interventions and Independence for Elders (LIFE) study manual, held in the NIA's Aging Research Biobank, uses a 4-metre walk, takes the faster of two walks to compute walking speed and requires the three parts to be done in a fixed order.

The LIFE manual assigns 0 points to a participant unable to perform a task and adds the walking, chair stand and balance scores into a summary that runs from 0 for the worst performers to 12 for the best. Timed results are converted to points through cut points, so two people a second apart can share a score while two people a fraction of a second apart can fall into different categories.

What the original cohort studies showed

The battery was built to predict, and its first validation was prognostic. In the 1994 paper, both self-reported disability items and the performance tests were independent predictors of short-term mortality and nursing home admission in multivariate analyses. Among people who reported almost no disability, the performance scores still separated a gradient of risk for both outcomes, which is the property that later made the SPPB attractive for identifying people at risk before disability appears.

The 1995 study in the New England Journal of Medicine followed 1,122 community-living people aged 71 and over who had no disability in activities of daily living at baseline. After adjustment for age, sex and chronic disease, those with the lowest scores were 4.2 to 4.9 times as likely to have disability in activities of daily living or mobility four years later as those with the highest scores, and those with intermediate scores were 1.6 to 1.8 times as likely. Guralnik and colleagues concluded that the tests might identify a preclinical stage at which intervention could prevent disability.

Later pooled evidence covers mortality. Pavasini and colleagues obtained standardised data from 17 studies with 16,534 participants, mean age 76 years, and reported in 2016 that an SPPB score below 10 was predictive of all-cause mortality. For a modeller, findings of this kind are what justify linking SPPB states to transition probabilities for death, institutionalisation or disability, although each link still needs its own source.

The SPPB as an eligibility rule and a trial endpoint

Trials of exercise and physical activity in older people often use the battery twice, once to select participants and again as the outcome. Selecting on a middle band of scores excludes people who are already severely limited and those at the top of the scale, where little improvement is possible.

The REACT (Retirement in Action) trial illustrates both roles. It randomised 777 people, mean age 77.6 years and mean baseline SPPB 7.37, to a group physical activity and behaviour maintenance programme or to control, and the economic evaluation notes that participants entered with an initial SPPB in the range 4 to 9. The primary outcome was change in SPPB score at 24 months, and the adjusted mean difference was 0.49 points (95 per cent confidence interval 0.06 to 0.92) in favour of the programme. Stathi and colleagues judged this to be just below their predefined clinically meaningful difference of 0.50 points, and the economic meaning of the difference depends on what a point predicts for later care use and quality of life, which the trial alone does not show.

Three routes linking SPPB scores with costs and QALYs

An SPPB result reaches an economic evaluation by one of three routes. Each route answers a different question, and the choice changes whether the result can be compared with a cost-effectiveness threshold.

Cost per person who improves

The simplest route treats improvement on the battery as a natural unit in a cost-effectiveness analysis. Farag and colleagues evaluated a home exercise programme alongside a randomised trial of 340 people aged 60 and over after a hospital admission. They reported an incremental cost of AUD 22,958 per extra person showing an improvement in mobility on the SPPB, AUD 19,020 per extra person reporting an improvement in health and AUD 77,403 per QALY. The mobility result reached an 80 per cent probability of cost-effectiveness at a willingness to pay of about AUD 48,000 per extra person improved, while no threshold gave that probability for the QALY outcome. The contrast shows how a programme can look good in physical performance terms and uncertain once the result is expressed in QALYs.

SPPB scores as the states of a decision model

A second route builds the model on the battery itself. The REACT economic evaluation by Snowsill and colleagues used a Markov model with a health state for each of the 13 possible SPPB scores, 0 to 12, and a state for death, to forecast mobility beyond the 24 months of trial follow-up. Costs and health-related quality of life were attached to the forecast mobility, the perspective was that of the NHS and personal social services, and costs and QALYs beyond 12 months were discounted at 3.5 per cent a year.

Within the trial, the programme was associated with a non-significant saving of GBP 103 and a significant gain of 0.040 QALYs (95 per cent confidence interval 0.009 to 0.071). Using the bootstrap percentile method on the trial data, the upper limit of the confidence interval for the ICER was estimated at GBP 17,000 per QALY. In the modelled subgroups the programme was least cost-effective for men aged 65, although it still dominated control, and most cost-effective for women aged 85. The authors noted that participants entered with an SPPB of 4 to 9, so transitions from scores outside that range rest on thinner evidence.

Collecting a preference-based measure alongside

The third route is design, not modelling. When a trial collects the EQ-5D or another preference-based measure at the same visits as the SPPB, QALYs come directly from the utility data and the SPPB serves as the clinical endpoint that explains them. A mapping function from SPPB scores to utilities would be needed only when no utility data were collected.

Illustrative worked example of two cost-effectiveness ratios

All numbers in this example are illustrative. A strength and balance programme for 1,000 older people costs GBP 600 a person more than usual care, and lower use of health and social care over two years offsets GBP 150 of that, so the incremental cost is GBP 450 a person. In the trial, 45 per cent of the programme group and 30 per cent of the control group improve by at least one SPPB point, and the mean QALY gain per person, from EQ-5D collected at the same visits, is 0.02.

1. Extra improvers. The difference in the proportion improving is 0.45 − 0.30 = 0.15, so the programme produces 0.15 extra improvers per person treated, or 150 among the 1,000.

2. Cost per extra improver.

$$\text{CER} = \frac{\Delta C}{\Delta p} = \frac{450}{0.15} = 3000$$

where $\text{CER}$ is the incremental cost per extra person improving by at least one SPPB point, $\Delta C$ is the incremental cost per person in GBP and $\Delta p$ is the difference in the proportion improving.

3. Cost per QALY. Dividing the same incremental cost by the QALY gain gives 450 ÷ 0.02 = 22,500, so the ICER is GBP 22,500 per QALY.

The two ratios answer different questions. GBP 3,000 per extra improver has no agreed benchmark, so a decision maker cannot tell from it alone whether the programme is good value. GBP 22,500 per QALY can be set against a threshold, but it depends on the EQ-5D data and not on the SPPB, which here explains the gain without valuing it. If the trial had not collected utilities, the analyst would need a model or a mapping to convert points into QALYs, and that step would carry its own uncertainty.

How the SPPB differs from nearby measures

The SPPB is one of several performance-based tests used in older people, and the boundaries matter when trials are pooled or a model borrows evidence from another study. The six-minute walk test records the distance walked in six minutes, a much longer task than the SPPB's short walk at usual pace. The timed up and go test combines rising, walking, turning and sitting into one timed sequence and reports seconds, not points.

Self-reported measures of disability, such as questionnaires about activities of daily living, record what a person says they can do, while the SPPB records what an assessor observes. Guralnik and colleagues found that the two kinds of measure each added independent prognostic information, so neither is a substitute for the other. A patient-reported outcome such as the EQ-5D mobility dimension is preference-weighted through a value set, while the SPPB is not. Frailty instruments often include gait speed or chair rises, but frailty is a wider syndrome and an SPPB score on its own is not a frailty diagnosis.

Limitations and common misreadings

Several features of the battery are easy to misread in an economic evaluation. Each one can change the size or the meaning of an estimated effect.

• Points are not utilities. The summary score is a sum of ordinal category scores, so a one-point change near the bottom of the scale need not mean the same change in health or care needs as a one-point change near the top.
• Ceiling at 12. A person scoring 12 cannot show improvement. Trials such as REACT that enrol only a middle band of scores avoid the ceiling, but their results do not transfer directly to fitter populations.
• Protocol variants. The original 8-foot walk and the later 4-metre walk give different times, and points come from cut points tied to a protocol, so pooled or borrowed SPPB data need the version checked.
• Different cut points for different purposes. REACT selected people scoring 4 to 9, while Pavasini and colleagues found a score below 10 predictive of mortality. A band chosen for eligibility should not be read as a clinical threshold for a model.
• Natural units without a benchmark. A cost per extra person improving on the SPPB, as in the study by Farag and colleagues, cannot be compared with a QALY threshold, and the same programme can look different once QALYs are used.
• Prediction is not causation. The cohort studies show that low scores predict disability, nursing home admission and death. A model that assumes raising someone's score lowers those risks by the amount the cohort gradient implies is making a causal assumption the cohort studies did not test.

Sources

• Farag I, Howard K, Hayes AJ, et al. Cost-effectiveness of a home-exercise program among older people after hospitalization. Journal of the American Medical Directors Association. 2015;16(6):490-496. PubMed 25703448. https://pubmed.ncbi.nlm.nih.gov/25703448/
• Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. Journal of Gerontology. 1994;49(2):M85-M94. doi:10.1093/geronj/49.2.M85
• Guralnik JM, Ferrucci L, Simonsick EM, Salive ME, Wallace RB. Lower-extremity function in persons over the age of 70 years as a predictor of subsequent disability. New England Journal of Medicine. 1995;332(9):556-561. doi:10.1056/NEJM199503023320902
• LIFE Study Investigators. Lifestyle Interventions and Independence for Elders (LIFE) Study, Manual of Procedures, Chapter 16 Physical Measurements, version 5.0, 20 February 2013. National Institute on Aging Aging Research Biobank. https://agingresearchbiobank.nia.nih.gov/studies/life/documents/download/Manual_of_Procedures_Main_Trial/Chapter_16_Physical_Measurements_v5.0.pdf/
• National Institute on Aging, Laboratory of Epidemiology and Population Sciences. Short Physical Performance Battery (SPPB). Web page, accessed October 2026. https://www.nia.nih.gov/research/labs/leps/short-physical-performance-battery-sppb
• Pavasini R, Guralnik J, Brown JC, et al. Short Physical Performance Battery and all-cause mortality: systematic review and meta-analysis. BMC Medicine. 2016;14:215. doi:10.1186/s12916-016-0763-7
• Snowsill TM, Stathi A, Green C, et al. Cost-effectiveness of a physical activity and behaviour maintenance programme on functional mobility decline in older adults: an economic evaluation of the REACT (Retirement in Action) trial. Lancet Public Health. 2022. doi:10.1016/S2468-2667(22)00030-5
• Stathi A, Greaves CJ, Thompson JL, et al. Effect of a physical activity and behaviour maintenance programme on functional mobility decline in older adults: the REACT (Retirement in Action) randomised controlled trial. Lancet Public Health. 2022. doi:10.1016/S2468-2667(22)00004-4