Hand grip strength and future hospitalisations in Māori and non-Māori New Zealand octogenarians: the LiLACS NZ cohort study

Previous research on the impact of chronic sarcopenia and low hand grip strength (HGS) on the hospitalisation of older people has shown those with low HGS are at greater risk of hospitalisation.^1,2 The incidence rate ratio (IRR) between the weakest and strongest quartiles was 1.56 (1.31–1.85).

A study of octogenarian men found that probable chronic sarcopenia (PCS) was associated with more time in hospital.³ The study found an unadjusted IRR for hospitalisation rate between those with PCS and those without of 1.46 (1.05–2.03).

The Second European Working Group on Sarcopenia in Older People (EWGSOP2) proposed using HGS rather than muscle mass to screen for PCS.⁴ This change would expedite chronic sarcopenia diagnoses in clinical practice, screening with a cheap and quick method, only testing the muscle mass and quality of those with low HGS.

Here, the association between PCS and HGS with hospitalisation was examined in the New Zealand context in Māori and non-Māori populations.

The impact of PCS and low HGS on any hospitalisation measure is important. More frequent and longer hospitalisations are measures of poor health and result in costly treatment. Using HGS to identify PCS early, particularly for Māori, could target interventions to reduce hospitalisations and promote health equity.

Objectives

This paper examined whether HGS and PCS were associated with future acute hospitalisations for community-dwelling Māori and non-Māori octogenarians in New Zealand.

Methods

The data analysed here are from LiLACS NZ (Life and Living in Advanced Age, a Cohort Study in New Zealand), a birth cohort study conducted by researchers from The University of Auckland that followed non-Māori from their 85th year and Māori in their 80s.⁵

The LiLACS NZ study recruited participants from the general population in the Bay of Plenty region of New Zealand between 2010 and 2011. There were six annual waves of interviews and physical examinations from 2010 to 2016. Both were usually conducted at the participant’s home or a dedicated assessment centre, occasionally at a primary healthcare clinic.

Written informed consent was requested from study participants to access routinely collected Ministry of Health – Manatū Hauora data.

All those living in the study area born in 1925, or self-identified as Māori and born between 1920 and 1930, were eligible to participate in the LiLACS NZ study. The widened age range for Māori enlarged the population pool for the Indigenous participants to that of the non-Māori group.⁵ The recruitment birth-year ranges were chosen to have a roughly 10% mortality rate per annum, based on Stats NZ Tatauranga Aotearoa’s life tables.⁶

Seven local primary healthcare providers conducted recruitment, interviews and physical assessments. These organisations used the electoral roll, their patient lists and advertisements to recruit participants. Each year after recruitment, participants were telephoned to arrange a follow-up interview and physical assessment, unless they had previously declined to participate further or had died.

The impact of PCS and low HGS on hospitalisation rate and total time in hospital were examined. Hospitalisation rate is reported here as hospitalisations per year and total time in hospital as days per year. Although these outcomes were related, they varied independently.

Hospitalisation data were from the National Minimum Data Set, collected by the Ministry of Health – Manatū Hauora and included admission and discharge dates and the facility to which the participant was admitted. The Ministry of Health – Manatū Hauora dataset included all admissions to medical facilities funded by the Ministry, including short emergency department visits and long-term stays in aged residential care (ARC). ARCs are generally known as rest homes or nursing homes. Visits to general practices and visits to facilities not funded by the Ministry were not included in the dataset. ARC admissions were excluded from the current analyses because the Ministry of Health – Manatū Hauora dataset did not include admissions not funded by the Ministry. Admissions to ARC were generally long-term and usually permanent; excluding them concentrates on short-term acute hospital treatment.

Muscle strength was assessed using HGS as part of the physical assessment. The participant squeezed the handgrip of the Takei GRIP-D dynamometer (Takei Scientific Instruments Co., Ltd, Tokyo, Japan) while standing with their arm at their side and their elbow extended fully three times with each hand. The highest of the six readings was used. If the participant could not stand for the test, it was performed seated with the arm in the same position as if standing. PCS derived from HGS was also tested as an explanatory variable. The cut-offs used are from the EWGSOP2 (≤27kg for men, ≤16kg for women).⁴

Participants’ age at their initial interview was controlled for in all models. Sex and ethnic grouping were self-defined in the initial interview. Hospitalisations before the HGS assessment were potentially confounding correlates of hospitalisations after the HGS assessment. Hospitalisation in the previous year was dichotomised into hospitalised or not and tested as a covariate.

Potentially fatal diseases and chronic conditions could be confounders, but most health conditions were too rare to be examined separately. Instead, the Multimorbidity Index (M3) score was tested as a covariate. The M3 score is derived from hospitalisation data, and its development team found that in New Zealand the M3 outperformed the Charlson and Elixhauser indices in predicting mortality.⁷ The M3’s creators recommend using either a 1- or 5-year history of hospitalisations; as LiLACS NZ had a relatively small sample size, the latter was selected.

The LiLACS NZ team attempted to recruit as many eligible individuals as possible within the survey area.⁸

The sample size for the LiLACS NZ study (approximately 500 Māori and 500 non-Māori) was chosen to detect a difference in mortality rate between different levels of nutrition risk or activities of daily living, with an expected overall mortality rate of 10% per annum.⁵ To increase recruitment, both full and core questionnaires were offered to participants, and a further physical assessment was optional.

Statistical methods

The hospitalisations used for analyses were divided between the previous year and the follow-up period. The previous year covered the 365 days preceding a participant’s enrolment in the study. The follow-up period for a participant was from study entry until death or 30 June 2024 (the last date of hospitalisations available). Hospitalisations during the follow-up period were the outcome; those from the previous year indicate a continuing pattern of admissions and were evaluated as a potentially confounding covariate to HGS and PCS. For simplicity, the previous years’ hospitalisations were dichotomised into hospitalised or not.

Generalised linear models were used to analyse the hospitalisation rate and days in hospital of participants during the follow-up period. As the distribution of this outcome was anticipated to be negative-binomial, a log-link was used, i.e., the natural logarithm of the expected number of hospitalisations was modelled as the outcome. The models were offset by the log of the follow-up period, as this varied between participants. The Pearson χ² statistic, divided by the degrees of freedom, was used to check each model for possible under- or over-dispersion.

A series of predictive models were constructed for each hospitalisation outcome for Māori and non-Māori, men and women separately. Men have much higher HGS than women, so separate models for each sex avoided the need for interaction terms in a single model. The relationship between HGS and hospitalisations could differ between Māori and non-Māori and would require independent age adjustments for the two groups. Separate models for each ethnicity avoided the need for interaction terms. First, either HGS or PCS as the sole explanatory variable, then controlling for age. Then, models controlling for age and previous hospitalisation, followed by models controlling for age and comorbidities, were tested. The exponential of the coefficients in the models was used to estimate the age and comorbidity adjusted incidence rate ratio (aIRR). The aIRR where HGS is the explanatory variable was the estimated incidence rate at any given HGS divided by the incidence rate of the HGS 1kg weaker at the same age. The aIRR where PCS is the explanatory variable was the estimated incidence rate of those with PCS divided by the incidence rate of those without PCS at the same age.

All analyses were performed using SAS® software, version 9.4 (TS1M8) for Windows (SAS Institute Inc., Cary, North Carolina, United States of America).

Hospitalisation rate

In the models of participants’ hospitalisation rates during the follow-up period, transfers of participants from one facility to another were counted as separate admissions, though they are within a continuous hospitalisation. (Internal transfers were already recorded as a single admission by the Ministry of Health – Manatū Hauora). Sensitivity analyses excluded transfers from the number of hospitalisations to ensure they did not impact results.

Total time in hospital

The number of days a study participant stayed in hospital during the follow-up period was modelled in the same manner as the hospitalisation rate. Ministry of Health – Manatū Hauora data record the length of stay as a number of nights; for analyses, this was changed to days by adding one to the total length of stay (transfers are recorded as separate admissions, but admission and discharge are on the same day). This change meant the minimum length of an admission was one rather than zero, distinguishing those discharged from hospital on the same day from those not hospitalised.

Other analyses

Missing data

Age, sex and ethnicity were known for all participants. Some participants did not have an initial HGS measurement. Models were constructed classifying participants as having, not having or unknown PCS to assess the impact of missing observations on results. Participants who did not wish their hospital records to be accessed by the researchers were excluded from all hospitalisation models (21 of those with HGS).

Hospitalisation quartile

The unadjusted hospitalisation rate models were re-run with HGS translated into sex- and ethnicity-specific quartiles to match the approach of a 2009 paper.¹ Like the earlier study, the IRR between the weakest and strongest quartiles and between those untested and the strongest quartiles were reported. As those untested were included, the explanatory variable was not treated as ordinal, but as a five-level categorical variable.

Results

Participants

The study recruited 56% of eligible Māori and 59% of eligible non-Māori in the recruitment area.⁹

Participants left the study due to personal preference, relocation outside the study area, illness or, most commonly, death.⁸ Leaving the study would have only affected hospitalisation data collection by the Ministry of Health – Manatū Hauora if the participant moved overseas, and this did not occur.

Descriptive data

Women comprised 54% of the participants included in the modelling and Māori made up 39%. The mean age of Māori was 82, younger than the non-Māori mean of 85 years. Thirty-four percent of study participants were hospitalised in the year before joining the study. Overall, a quarter of those tested had PCS (Table 1).

Only two-thirds of participants felt willing and able to be physically assessed, so 67% had their HGS measured (60% of Māori, 72% of non-Māori). Hospitalisation numbers were available for those who granted permission to access their Ministry of Health – Manatū Hauora records, 94% of study participants (90% of Māori, 97% of non-Māori).

Outcome data

Most study participants (93%) were admitted to hospital at some point during follow-up (Table 1), and a minority (44%) were admitted more than once a year (Table 2). During follow-up, most (66%) participants spent less than a week in hospital per year.

Main results

Comparing those with and without PCS, those without PCS had better outcomes on average (Table 2). Those without PCS had a lower hospitalisation rate and spent less time in hospital.

Hospitalisation rate

HGS had a significant inverse association with the hospitalisation rate for Māori women and non-Māori men, controlling for age, previous hospitalisation or chronic conditions (Table 3). For Māori women, the aIRR per 1kg increase in HGS was 0.968 (95% confidence interval; 0.942–0.995). For non-Māori men, the aIRR was 0.970 (0.951–0.991). All the HGS models indicated that greater strength was associated with fewer hospitalisations (Table 3).

The age-adjusted models for hospitalisation rate showed PCS to be a significant explanatory variable for non-Māori men only (Table 3). PCS retained a significant association with the hospitalisation rate for non-Māori men when controlling for previous hospitalisation or chronic conditions (Table 3). The coefficients in all the PCS models indicated a higher hospitalisation rate in those with PCS. This difference is noticeable in the summary numbers, with a higher proportion of those with PCS being admitted to hospital more than once a year (Table 2).

Total time in hospital

HGS had a significant association with the total time in hospital for Māori and non-Māori men, controlling for age and previous hospitalisation (Table 3). For non-Māori men, but not Māori men, this association remained statistically significant when controlling for chronic conditions. However, all the HGS models indicated that greater strength was associated with less time in hospital.

The age-adjusted models for total time in hospital showed PCS to be a significant explanatory variable for Māori and non-Māori men. PCS retained a significant association with the total time in hospital when controlling for previous hospitalisation or chronic conditions for non-Māori men (Table 3). The coefficients in all the PCS models indicated more time in hospital for those with PCS. The summary numbers show a higher proportion of those with PCS having a total time in hospital longer than 1 day per year (Table 2).

Other analyses

Missing data

For Māori women, those without an HGS test had a significantly higher hospitalisation rate and total time in hospital. For Māori men, the modelled time for those with unknown status lay between the modelled time for those with and without PCS.

Hospitalisation quartile

In the hospitalisation rate models with HGS quartiles as explanatory variables, the IRRs between the weakest and strongest quartiles were 1.85 (1.27–2.69) for Māori women, 1.20 (0.69–2.09) for Māori men, 1.10 (0.80–1.53) for non-Māori women and 1.83 (1.32–2.53) for non-Māori men. The IRRs between those not assessed and the strongest quartiles were 2.56 (1.83–3.59) for Māori women, 1.23 (0.79–1.93) for Māori men, 0.92 (0.68–1.23) for non-Māori women and 1.29 (0.92–1.81) for non-Māori men.

Discussion

Key results

Lower HGS was associated with more frequent future hospitalisations for Māori women, but not the total time in hospital. Both lower HGS and PCS were associated in Māori men with total time in hospital. Existing medical conditions, which could impact both HGS and hospitalisation, were more associated with total time in hospital. Neither lower HGS nor PCS in non-Māori women were associated with hospitalisation frequency or duration. Both lower HGS and having PCS in non-Māori men were associated with more frequent hospitalisations and total time in hospital. Why these relationships were not significant for non-Māori women is unclear. Since some of the models for Māori showed a significant relationship, it is possible that a larger sample of Māori could have shown statistical significance in the other models.

Limitations

Because multiple tests were explored and there was a false-positive (Type I error) rate of 5%, readers should treat results with caution. Conversely, the possibility of a false negative (Type II error) due to small sample sizes should be considered.

Data source

Proportionately, fewer women and ARC residents were in the study than were in the general population.⁸ As an epidemiological study, LiLACS NZ tried to accommodate those living with chronic illnesses,⁹ but individuals declined to participate because of poor health. The missing data model for Māori women suggests that a disproportionate number of those who felt well enough to take part in the study did not feel well enough to have their HGS assessed.

Studies that found significant relationships between hospitalisations and HGS or sarcopenia had larger sample sizes.^1–3 The LiLACS NZ models have similar coefficients regardless of significance. Perhaps significant relationships could have been observed with a larger cohort.

Outcome

Hospitalisations were restricted to those funded by the Ministry of Health – Manatū Hauora, excluding self-funded hospital admissions. In the study area, public hospitals had 661 beds compared to 73 beds in private hospitals (excluding a maternity hospital and a hospice), so the volume of private hospitalisations was likely less than 10% of the total.^10,11

Explanatory variables

HGS

There is no universally agreed-upon method for measuring HGS; slightly higher readings are observed when subjects stand with their arm at their side than when sitting down with their elbow at 90 degrees.¹² Readings can vary between equipment from different manufacturers.¹³ The EWGSOP2 cut-offs are based on a paper that pooled data from 12 studies; most used the JAMAR dynamometer while seated, only the Newcastle 85+ study used the Takei device while standing as LiLACS NZ did.¹⁴ The authors of that paper checked but did not find it necessary to adjust for the differing methods and devices when pooling the data. The LiLACS NZ study employed the same dynamometer and methods with the same age group as Newcastle 85+, so the results should have a similar level of compatibility with other studies.

PCS

PCS had a poorer association with future frequent hospitalisations and total time in hospital than HGS. This disadvantage was expected when dichotomising the measure, but it defined the group where a possible intervention could be targeted.

The EWGSOP2 team acknowledged that their selection of cut-offs for HGS and other measures was arbitrary and a topic of further research. The EWGSOP2 procedure for defining chronic sarcopenia identifies probable sarcopenia using muscle strength, and then poor muscle mass confirms the diagnosis.⁴ There are reasons for low strength other than sarcopenia, such as degraded neuromuscular signals.¹⁵ However, in the LiLACS NZ study, few participants had dementia, and excluding those with a Modified Mini-Mental State Examination (3MS) score below 70 had a negligible effect on the models.

Interpretation

The findings here are consistent with previous research. A systematic review in 2016 found only one study that looked at the relationship between HGS and subsequent hospitalisation.² That study, which involved Americans in their 70s, reported an increased hospitalisation rate in the follow-up period for those in the lowest quartile of HGS compared to the highest quartile.¹ Over 44% of participants were not admitted to hospital despite the follow-up period averaging 4.7 years. Quartiles were sex- and ethnicity-specific, so someone with a HGS of 31kg would be classed as belonging to the strongest quartile if they were a woman, whereas they would be classed as in the weakest quartile if they were a man. The IRR between the weakest and strongest quartiles was 1.56 (1.31–1.85). The IRR between those untested and the strongest quartile was 1.61 (1.00–2.59). Despite the differing age ranges and quartile cut-offs, the IRRs comparing the weakest and the untested with the strongest were quite similar between the two studies.

A study of octogenarian men found an unadjusted IRR for hospitalisation rate between PCS and no PCS of 1.46 (1.05–2.03), which is within the range of aIRRs observed in the LiLACS NZ study; therefore, the two studies are reasonably consistent.³

Both HGS and PCS were associated with future hospitalisations for non-Māori men, consistent with previous research.^1–3 Both HGS and PCS were associated with future time in hospital for Māori and non-Māori men, consistent with previous research.³ The LiLACS NZ data, even when the relationships were not statistically significant, were consistent with previous research.

Conclusions

The associations observed suggest that HGS could be used for screening in a primary care setting to identify individuals at risk of more frequent and prolonged hospitalisations. Further research could explore the utility of HGS as a screening tool for targeting treatments for chronic sarcopenia early. Besides possible cost benefits to the hospital system, screening older Māori people to reduce their rate of chronic sarcopenia could reduce inequities in hospitalisation outcomes.

Various approaches to treating chronic sarcopenia have been investigated in other studies. These treatments include exercise,¹⁶ reducing drinking and smoking,¹⁷ diet¹⁸ and dietary supplements.¹⁹ Whether these treatments alone, or in combination, would be beneficial in reducing hospitalisations could also be the subject of further research.

View Table 1–3.