Physical and mental functioning trajectory classes among older adults and their association with specialized healthcare use

The participants of the present study are a subsample of the Helsinki Birth Cohort Study of 8760 individuals born in Helsinki University Hospital between 1934 and 1944 [15]. A random sample of 2003 individuals participated in clinical examinations between the years 2001 and 2004. Follow-up examinations were performed between the years 2011 and 2013 and 2017 and 2018 as seen in the flowchart in Supplementary Fig. 1. Physical and mental functioning were assessed with the validated 36-Item Short Form Health Survey [16, 17] at all three study visits for those who participated (Supplementary Figure 1). Specialized healthcare use was obtained from a national specialized healthcare register starting from the first clinical examination date and ending at the end of the year 2017. Overall, 1999 participants had data on physical and mental functioning from at least one clinical visit to determine their physical and mental functioning. Eight individuals were excluded due to death in less than a year after the first clinical examination. The final study population consisted of 1991 individuals. All study participants gave a written informed consent before participating in any clinical procedures. The clinical study was performed in accordance with the Declaration of Helsinki and approved by the Coordinating Ethics Committee of the Hospital District of Helsinki and Uusimaa (THL/101/6.02.00/2019).

Physical and mental functioning were assessed with the Finnish validated version of the 36-Item Short Form Health Survey Version 1.0 (SF-36), which measures self-reported general health and wellbeing [17]. The SF-36 consists of eight sub-domains which each include 2–10 items. The questionnaire has shown to be valid and practical instrument for older community dwelling older adults [18], and unidimensionality and reliability of the sub-scales have been proved among older adults [19]. The sub-domains are fundamentally associated with healthy functioning and are named as physical functioning (10 items), role limitation due to physical problems (4 items), bodily pain (2 items), general health perception (5 items), vitality (4 items), social functioning (2 items), role limitation due to emotional problems (3 items), and mental health (5 items). Each item was coded to range from 0 to 100 with higher scores indicating better functioning or well-being. Sub-domain scores were formed by calculating means of the items, and coded missing if less than half of the items were available. The scores of the sub-domains were then standardized [20] using the means and standard deviations of a US reference population from 1990, and then weighted using factor score coefficients from the US population. Next, the sub-domain scores were aggregated to create physical and mental component scores. Finally, both component scores were standardized using mean of 50 and standard deviation of 10 [21]. The same research assistant nurses carried out the procedures at all three clinical examinations. The functioning scores were further divided by 10 for analysis in mixture models.

Healthcare utilization data were retrieved from the Care Register of Health Care, a national specialized healthcare register maintained by the Finnish Institute for Health and Welfare [22]. The register includes data on outpatient and inpatient healthcare visits from all hospitals in Finland. The data were retrieved starting from the first clinical examination date organized during 2001–2004 and continued until the end of the year 2017 or time of death where applicable. Emergency visits and first and follow-up outpatient visits were directly extracted from the register whereas hospital days were calculated by subtracting the admission date from the discharge date. If inpatient care exceeded 365 days, it was no longer treated as acute care and coded as 365 days in the dataset. Visits or hospital days were further summed for each study participant. First outpatient visits included all visits where a patient care relationship started with a new referral to a specific specialty while follow-up outpatient visits included all visits regarding the check-ups of the previously examined chronic diseases or injuries.

Covariates were assessed in the first clinical examination. Self-reported physician diagnosed chronic diseases or symptoms included myocardial infarction, angina pectoris, congestive heart failure, claudication, osteoporosis, stroke, depression, asthma, and emphysema. Self-reported chronic diseases were divided into three categories including ‘no diseases’, ‘one chronic disease’, and ‘two or more chronic diseases’. Frequency of smoking was assessed with a questionnaire and categorized into ‘never’, ‘former’, and ‘current smoker’. Self-reported alcohol consumption was categorized into ‘does not consume’, ‘two times a month at most’, and ‘three times a month or more’. Information on the participants’ highest educational attainment was obtained from Statistics Finland [23] in the year 2000 and divided into four groups: ‘basic or less or unknown’, ‘upper secondary’, ‘lower tertiary including polytechnic, vocational, and bachelor’s degree’, and ‘upper tertiary including master’s degree or higher’. Information on leisure time physical activity (LTPA) was assessed with the Kuopio Ischemic Heart Disease Risk Factor Study questionnaire [24]. Participants were asked about the duration, frequency, and intensity of LTPA during the past 12 months. Based on activity-specific metabolic equivalent (MET) values for each intensity grade, total leisure-time physical activity was calculated, and the results were expressed as METhours per week.

To form physical and mental functioning trajectory classes, latent growth mixture models (LGMM) were fitted to the SF-36 summary component scores available at the clinical examinations in 2001–2004, 2011–2013 and 2017–2018. The analyses were performed separately for women and men with Mplus version 7 (1998–2015). Assuming non-response was missing at random, we used LGMM with Full Information Maximum Likelihood [25] to capture unobserved latent classes using all available data with similar physical and mental functioning trajectories, but which were relatively distinct across the latent classes over follow-up time. Within the latent classes, the component scores were used as outcomes in a growth model with time indexed through individually varying observation times corresponding to the chronological age of participants. Each latent class had their own growth parameters, intercept (indicating participant’s level of physical or mental functioning scores) and slope (indicating participant’s rate of linear changes in physical or mental functioning scores over time). For the classes, the average trajectory was based on the means of the intercept and slope parameters calculated at age means at each time point. Final classification of the participants was assigned by the highest group membership posterior probability which was based on data from each participant’s observed trajectory. The optimal number of latent classes was determined using the Bayesian Information Criterion (BIC), with lower values indicating a better fit of the model, and clarity of the participants’ classification. Clarity of classification into trajectory classes was assessed with percentage of individuals falling into latent classes based on membership posterior probabilities (high probabilities indicates clear separation of trajectory classes) and summarized by model entropy (an aggregate function of posterior probabilities), which ranges between zero and one, with higher values one indicating better class separation. Average trajectory classes were named based on the starting level and the slope of the graphical average trajectory: trajectory classes with better functioning between the numbers were named ‘high’ and the lower average trajectory classes ‘intermediate’ due to graphical presentation approximately in the middle of the chart. Trajectory classes with no or little slope were named ‘stable’ since a slight decrease in functioning was considered to be related to normal aging. Trajectories with more declining slopes were named ‘declining.

Means and standard deviations for continuous variables and proportions for categorical variables were calculated to compare baseline characteristics. Statistical differences in baseline characteristics were examined with the independent sample t-test, the Mann–Whitney U test, or the chi-square test. We used negative binomial regression models to examine the associations between physical and mental functioning trajectory classes and specialized healthcare use. Two models were created: Model 1 was adjusted for age, and Model 2 additionally for LTPA, education, number of chronic diseases, smoking, and alcohol consumption. The analyses were performed separately for men and women, and individual exposure times (offsets) were adjusted for the models. The results are shown as incidence rate ratios (IRRs). Significance level was set at 0.05, and p-values were not corrected for multiple testing. Further, negative binomial regression models were used to perform sensitivity analyses where we examined the associations between specialized healthcare use and physical and mental functioning trajectory classes with membership probability of 0.85 or higher. The analyses were performed in SPSS 26.0 for Windows (version 26.0, 1989–2020, SPSS Inc., Armonk, NY, USA).

Figures 1 and 2 show physical and mental functioning trajectories in the two classes identified for men and women. The two physical functioning trajectory classes for an average participant among men and women were named high declining (65.6% of men, 55.5% of women) and intermediate declining (34.4 of men, 44.5% of women). As presented in Figs. 1 and 2, two mental functioning average trajectory classes were identified for both men and women: high stable (77.5% of men and 65.9% of women) and intermediate stable (22.5% of men and 34.1% of women). We refer to these trajectory classes as high and intermediate hereon in this manuscript. Supplementary Figs. 2 and 3 in the Supplement show all observed individual physical and mental functioning trajectories by sex.

Table 1 presents the characteristics of the study population. Men were more likely to have higher education, were more likely ex- or current smokers, were characterised by higher alcohol consumption, and had slightly higher annual utilization rates of emergency visits, first outpatient visits, and hospital days compared to women (Table 1). There were no differences between sexes in age, mean number of chronic diseases, and LTPA (Table 1). Supplementary Tables 1 and 2 in the Supplement show the characteristics of the study population according to the physical and mental functioning trajectory classes and sex.

Table 2 displays the results of regression analyses regarding physical functioning trajectory classes and specialized healthcare service use. Both men and women in the intermediate physical functioning trajectory class had higher utilization rates of any specialized healthcare service than those in the high trajectory class (Table 2). After all adjustments, the IRRs of both sexes were similar (Table 2).

Table 3 shows the IRRs for specialized healthcare service use by mental functioning trajectory classes. Men and women classified into the intermediate trajectory class utilized first outpatient visits at higher rates in comparison to the high trajectory class. The utilization rates of first outpatient visits in the intermediate trajectory class were similar among both sexes (Table 3). In fully-adjusted models, no other statistically significant associations were observed (Table 3). Overall, the IRRs for specialized healthcare service use by mental functioning trajectory classes were smaller compared to the IRRs of physical functioning trajectory classes. The results were further confirmed in sensitivity analyses as seen in Supplementary Tables 3 and 4 in the Supplement.