Pain and disability transitions among older Americans: The role of education

Abstract

Previous literature has rarely examined the role of pain in the process of disablement. We investigate how pain associates with disability transitions among older adults, using educational attainment as a moderator. Data are from the National Health and Aging Trends Study, N=6,357; 33,201 one-year transitions between 2010–2020. We estimate multinomial logistic models predicting incidence or onset of and recovery from functional limitation and disability. Results show pain significantly predicts functional limitation and disability onset one year after a baseline observation, and decreases odds of recovery from functional limitation or disability. Contrary to expectations, higher education does not buffer the association of pain in onset of disability, but supporting expectations, it facilitates recovery from functional limitation or disability among those with pain. The analysis implicates pain as having a key role in the disablement process and suggests that education may moderate this with respect to coping with and subsequently recovering from disability.

1. Introduction

Disability rates in the U.S. have increased significantly in recent decades.^{9, 23} This is considered to be a function of a variety of factors, including population aging, increasing chronic disease prevalence, increasing psychological stress, and economic distress.^{5, 49} Higher disability prevalence has resulted in an urgent need to identify disability’s developmental process, its drivers, and groups most vulnerable. Despite a connection between the two, pain has been greatly under-studied as a factor in disablement.^{2, 7, 51} In this study, we contribute to this understanding by examining pain’s association with incidence or onset in and recovery from disability as well as education’s moderating role in these processes.

The “disablement process” advanced by Verbrugge and Jette⁴⁶ is one way of theoretically understanding the development of disability. The process suggests disability is the outcome of a set of linked factors, beginning with pathology (i.e., a medically labelled disease or condition) that causes impairment (i.e., consequential abnormalities including pain), which may develop into a functional limitation (such as difficulty walking or climbing stairs), which in turn leads to a disability, defined using Activities of Daily Living (ADLs) such as dressing and bathing.²⁴ Pain, therefore, in this model, has a direct impact on functional limitation and an indirect impact on disability. Substantial empirical evidence links pain to both functional limitation^{7, 36, 37} and disability.^{2, 25, 36, 39} Though the model of Verbrugge and Jette⁴⁶ does not say much about recovery, research suggests short-term movement in and out of disability is common for older adults, especially those newly disabled and with less severe forms of disability.^{8, 22, 28} But, the role of pain in recovery from disability is not well understood.

Since disability, expressed using an ADL operationalization, is defined as an inability to do activities necessary for survival within a specific physical and social environment, the conceptualization also considers that predisposing factors moderate the process by shaping how people are exposed to and interact with their environment. There is good deal of evidence implicating education as a determinant of both pain and disability and a factor in decreasing the risk of pain and disability.^{6, 12, 21, 28, 30, 31, 35, 40, 45} Therefore, education is a critical predisposing characteristic. For instance, higher education is associated with economic standing and stability, a better understanding of epidemiological processes, access to personal care, assistive device use, and social and psychological resources, and knowledge about healthy lifestyles.^{1, 10, 16, 20, 27, 34, 50} Empirical evidence indicates education is associated with development of functional limitation,^{12, 19, 29} suggesting a potential protective effect in disablement. Subsequently, education is likely to impact the degree to which pain manifests into a disability, and by helping to exploit salutary coping strategies, the degree to which pain hampers recovery from disability.

Using 10-year longitudinal panel data on late-life disability among older U.S. adults, we monitor one-year disability transitions to assess the degree to which pain and education, independently and together, are implicated in the disablement process. Drawing on the Verbrugge and Jette conceptualization,⁴⁶ the analysis considers the possibility that functional limitation is an intermediary step between pain and disability. We examine the processes of developing disability and recovery from disability, since determinants of recovery may not necessarily be the same as determinants of onset.⁸ We consider four hypotheses: pain increases the probability that someone without a functional limitation or disability at baseline with have a functional limitation or disability one-year later (H1). Pain decreases the chances of recovering from functional limitation or disability (H2). Education buffers the impact of pain on functional limitation and disability onset (H3). Education facilitates recovery from functional limitation or disability among those with pain (H4).

2. Methods

Data

The data are drawn from the National Health and Aging Trends Study (NHATS), which has as a primary focus to study disability and quality of life at older ages.¹⁵ NHATS uses a longitudinal panel design that collects data annually, by way of survey, from a nationally representative sample of Medicare beneficiaries aged 65 and over in the U.S. Our analysis focuses on the initial cohort to enter NHATS; a cohort first surveyed in 2011 and followed-up until 2020. This cohort was selected with a sampling design that used geographic clustering across strata, Primary Sampling Units (counties), and zip codes. Persons aged 90 and over and non-Hispanic black individuals were oversampled.^14,15 The response rate in 2011 was 70.9%, and the sample size was 8,245. The response rates for follow-ups were above 85%.¹⁵ We use all data waves between 2011 and 2020. Both direct and proxy interviews are included. NHATS has been granted waivers from properly constituted Institutional Review Board and informed consent from the respondents or their proxy is obtained.³²

Our analysis is based on one-year functional limitation/disability transitions. That is, our data is constructed using functional limitation and disability measures from pairs of adjacent survey waves. Therefore, we code a baseline functional limitation/disability status for a respondent in any survey wave and a follow-up functional limitation/disability status, or being deceased, in the following survey wave. In this way, any individual observed for at least two contiguous waves is included in the analysis, but individuals can contribute multiple transition data points if they remain in the data for at least three contiguous waves. Among the total sample of 8,245, 636 respondents living in nursing homes are excluded from the initial sample person interview but those moving into nursing homes after the initial wave of data collection are retained. We further removed 114 lacking information on covariates including education, race, and marital status, and 1,138 respondents who do not contribute to any transition due to missing data at either baseline or follow-up status. The final dataset contains 33,201 valid transitions from 6,357 respondents such that each contributes five valid transitions on average.

Measures

As in the disablement process conceptualization, we separate functional limitation from disability, consider functional limitation as a mediating event, and identify transitions from a baseline functional limitation/disability status to a follow-up functional limitation/disability status or death. Functional limitation is measured with six questionnaire items, including whether the respondent can complete six activities: walk three blocks, climb a flight of 10 stairs, lift 10 pounds, bend over, reach overhead, and grasp small objects. Those that are not able to complete at least one of these six tasks are coded as having a functional limitation. Disability is based on Activities of Daily Living introduced by Katz et al.²⁴ Seven tasks for daily survival are asked about in the survey: going outside home/building, getting around inside home/building, getting out of bed, eating, getting cleaned up, using the toilet, and getting dressed. If the respondent has no difficulties completing all tasks, either because they have the ability to do it themselves, they have the ability to do it with the use of assistive technology, or because someone does the task for them, they are categorized as having “no disability.” The NHATS tracker file provides information on the year and date of death for individuals deceased by follow-up.

Pain status is obtained by a survey question that asks “in the last month, have you been bothered by pain.” The responses are either yes or no and therefore the variable is dichotomous. Education is coded into three groups as follows: less than complete high school; complete high school but no college degree (i.e., complete high school, or some colleges or associate’s degree); complete college degree.

The analysis also adjusts models for age, sex, race/ethnicity, marital status, proxy responses, community residence, and baseline year. Age groups are coded as 65–69, 70–74,75–79, 80–84, 85–89, and 90 and above. Females are coded as 1 and males as 0. Racial/ethnic groups are non-Hispanic whites (whites hereafter), non-Hispanics blacks (blacks hereafter), non-Hispanic others (others hereafter), and Hispanics. Marital status is categorized as married or living with a partner, separated/divorced, widowed, and never married. The indicators for proxy respondents, community respondents (not residential care or nursing home), and the baseline year of the transition are also included.

Analytical strategy

Figure 1 displays the transition probabilities (which are equivalent to the percentages moving from state to state, from baseline to follow-up) estimated in this analysis, which pairs three baseline states with four follow-up states. The figure indicates that at baseline an individual can be coded as being: without functional limitation or disability, having a functional limitation but not a disability, or having a disability. At follow-up an individual is coded as being in one of these states or deceased. We stratify the sample into baseline groups and examine transition models from each baseline state, with the sum of probabilities from any baseline state to the four possible outcome states being equal to 1 (or the sum of percentages equal to 100). Because the outcomes are unordered, we use multinomial logistic regression models to estimate the probabilities. Transforming from the equation of multinomial regression modelling

$$\text{ln}\left(\frac{\text{Pr}\left(Y_i=k\right)}{\text{Pr}\left(Y_i=1\right)}\right)=X_i{\beta }_k^{\text{'}}$$

where k refers to categories of outcomes (k=2, 3, 4) with the 1^st outcome as the reference, the probabilities of four outcomes would be:

$$\text{Pr}\left(Y_i=k\right)=\frac{e^{X_i{\beta }_k^{\text{'}}}}{1+\sum e^{X_i{\beta }_k^{\text{'}}}}$$

where k=2, 3, 4, and

$$\text{Pr}\left(Y_i=1\right)=\frac{1}{1+\sum e^{X_i{\beta }_k^{\text{'}}}}$$

for the base outcome. These transition probabilities are considered to be a function of pain status at baseline, education level, the interaction between pain and education, and other covariates including age, sex, race/ethnicity, marital status, indicators for proxy and community responses, and the year of baseline. The interaction accounts for the possible moderating impact of education. So, for instance, there will be a probability of remaining without functional limitation or disability at both baseline and follow-up (i.e. P₁₁) for those without pain at baseline across three educational levels and for those with pain at baseline across three educational levels, adjusting for other covariates. This means that we estimate an array of six separate probabilities across combinations of pain and educational level. We follow the NHATS technical document¹⁴ to account for the geographic clustering sampling design, the differential selection probabilities and nonresponse. The regression models are adjusted for the stratum variable and sampling weights (which are the analytical weights for the follow-up year of each transition), and, since each individual in the data can contribute multiple transition observations, individual-level clustering. After estimating the multinomial regression models, predicted percentages of four follow-up states across baseline states, pain, and educational levels, are calculated using the “margins” procedure in STATA, with other covariates set at mean values.

Figure 1.

Estimated transition probabilities

Table 1 shows descriptive statistics for the total sample and samples with different baseline status. The baseline sample sizes are 14,531 transitions without functional limitation or disability, 4,064 with functional limitation only, and 14,606 with disability. People with functional limitation or disability are more likely to have pain and be older, less educated, females, racial/ethnic minorities, widowed/never married, have proxy interview, and be institutionalized compared to those without functional limitation or disability.

Table 1.

Weighted characteristics of the total sample and samples with different baseline status^a

		Baseline status
	Total Sample	Without	FL only	Disability
Baseline status
Without (%)	49.6
FL only (%)	11.3
Disability (%)	39.1
Pain (%)	55.5	42.1	59.1	71.4
Age
65–69 (%)	9.5	12.6	6.3	6.4
70–74 (%)	27.4	34.3	21.4	20.4
75–79 (%)	26.1	28.5	26.1	23.1
80–84 (%)	18.5	15.2	22.5	21.5
85–89 (%)	11.7	7.2	15.6	16.4
90 and above (%)	6.8	2.2	8.1	12.2
Education
Less than complete high school (%)	19.5	13.1	24.7	26.0
Complete high school but no college degree (%)	52.2	50.5	59.3	52.4
College degree (%)	28.3	36.4	16.0	21.6
Female (%)	56.8	49.8	67.4	62.7
Race/ethnicity
Whites (%)	81.8	86.3	75.7	77.8
Blacks (%)	8.3	6.1	9.9	10.6
Others (%)	3.5	3.3	5.0	3.3
Hispanics (%)	6.4	4.3	9.4	8.3
Marital status
Married or living with a partner (%)	53.4	62.9	39.8	45.2
Separated/divorced (%)	12.5	12.1	14.8	12.4
Widowed (%)	30.4	22.0	40.6	38.2
Never married (%)	3.7	3.0	4.8	4.2
Proxy (%)	4.8	0.7	1.6	11.0
Community (%)	93.1	97.2	93.0	87.8
n	33,201	14,531	4,064	14,606

^aWithout refers to without functional limitation or disability and FL refers to functional limitation.

3. Results

Table 2 shows weighted percentages of functional limitation/disability status at follow-up for the total sample and by functional limitation/disability and pain status at baseline. A t-test determines if the difference in corresponding percentages between those with and without pain is statistically significant. At follow-up, 46.0% are without functional limitation or disability, 10.8% have functional limitation only, 38.3% have a disability and 4.9% are deceased. Follow-up status is highly associated with baseline status, although there is a fair amount of movement as well across states. For instance, for the total sample, among those without functional limitation or disability at baseline, 6.8% have a functional limitation only, 14.9% have a disability, and 1.6% have died by follow-up. There is also the possibility of improving status. For instance, among the total sample with a disability at baseline, 13.7% are without functional limitation or disability at follow-up and 8.3% have functional limitation only.

Table 2.

Weighted percentages of follow-up status by baseline status and pain^b

Baseline status	Follow-up status	Total	With pain at baseline	Without pain at baselinec	Ratiod
Total sample	Without (%)	46.0	36.5	57.8***	0.63
	FL only (%)	10.8	11.7	9.7***	1.21
	Disability (%)	38.3	46.6	27.9***	1.67
	Died (%)	4.9	5.2	4.6**	1.14
	unweighted n	33,201	18,431	14,770
Without	Without (%)	76.7	72.6	79.7***	0.91
	FL only (%)	6.8	8.0	5.9***	1.35
	Disability (%)	14.9	18.0	12.7***	1.42
	Died (%)	1.6	1.4	1.7	0.83
	unweighted n	14,531	5,931	8,600
FL only	Without (%)	22.2	20.7	24.4*	0.85
	FL only (%)	37.6	37.8	37.3	1.01
	Disability (%)	35.2	37.5	31.9**	1.17
	Died (%)	5.0	4.0	6.4**	0.63
	unweighted n	4,064	2,350	1,714
Disability	Without (%)	13.7	13.1	15.2*	0.86
	FL only (%)	8.3	8.3	8.2	1.02
	Disability (%)	68.8	70.2	65.4***	1.07
	Died (%)	9.2	8.4	11.2***	0.75
	unweighted n	14,606	10,150	4,456

^bWithout refers to without functional limitation or disability and FL refers to functional limitation.

^cThe significance for t-test results is shown.

⁺p<0.1,

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

^dThe ratio of percentages being in a state for those with pain relative to those without.

Pain status at baseline is associated with transitions across all baseline statuses and therefore pain is implicated in the disablement process, at least when unadjusted for other covariates. For instance, among those without functional limitation or disability at baseline, those with pain are less likely to remain in that state one year later (72.6% versus 79.7%), and more likely to report functional limitations only (8.0% versus 5.9%) and disability (18.0% versus 12.7%). These differences are all statistically significant. Among those with functional limitation only at baseline, the percentage improving to being without functional limitation or disability at follow-up is lower (20.7% versus 24.4%), the percentage having a disability is greater (37.5% versus 31.9%) for those with pain, and these differences are also statistically meaningful. Among those with a disability, the percentage remaining with a disability is greater (70.2% versus 65.4%) and the percentage improving to being without functional limitation or disability is lower (13.1% versus 15.2%) for those with pain. Differences are again statistically significant.

A convenient way of summarizing the association between pain and follow-up status is to look at the ratio of percentages being in a state for those with pain relative to those without. A ratio lower than 1.00 indicates the chances of being in that state at follow-up are lower for those with pain, while a ratio higher than 1.00 indicates the opposite. The ratios of 0.85 and 0.86, which are seen for being without functional limitation or disability among those with functional limitation only or with disability at baseline, suggest quite a strong negative impact of pain on the likelihood of improving status. Ratios can also be interpreted as percentage differences such that those with a functional limitation and without pain at baseline are 15% more likely to improve to being without functional limitation or disability at follow-up, and those with disability and without pain at baseline are 14% more likely to improve to being without functional limitation or disability at follow-up. Similarly, a ratio of 1.42 for having a disability among those without functional limitation or disability at baseline indicates a strong positive impact of pain on the incidence of disability. Those with pain are 42% more likely to have disability at follow-up in comparison to those without pain.

The percentage dying presents somewhat of an anomaly. While the percentage dying is higher for those disabled than those not, within disablement states the percentage dying is higher for those without pain. However, this anomaly no longer shows up in multinomial regression models when other covariates are controlled, as seen below.

Table 3 presents results predicting functional limitation/disability status at follow-up by pain, education, and other covariates, at baseline. The contrast category for each of these regressions is the state that represents stability in status over time. 95% confidence intervals of coefficients are shown in supplemental Table S1. We only include interaction effects between pain and education when these effects are statistically significant. Both Model 2 and Model 3 have statistically significant interactions.

Table 3.

Multinomial logistic regression models (shown in log odds ratios) for the follow-up status based on three sub-samples^e

	Model 1 (base outcome: Without)			Model 2 (base outcome: FL only)			Model 3 (base outcome: Disability)
	FL only	Disability	Died	Without	Disability	Died	Without	FL only	Died
Pain	0.445***	0.498***	0.044	−0.712**	0.111	−0.411	−0.623***	−0.463**	−0.059
Less than high school (ref.)
Complete high school but no college degree	−0.378**	−0.225*	−0.223	−0.467*	−0.169	−0.396	0.142	−0.076	0.032
College degree	−1.177***	−0.291**	−1.090***	−0.176	−0.077	−0.438	0.536**	−0.043	−0.113
Pain X Complete high school but no college degree				0.593*	0.058	−0.014	0.150	0.490*	−0.017
Pain X College degree				0.828*	0.278	0.278	0.265	0.177	−0.187
65–69 (ref.)
70–74	0.381*	0.089	1.090**	−0.016	0.169	1.204+	−0.093	0.096	−0.116
75–79	0.540**	0.310**	1.234**	−0.349	0.314	1.331+	−0.503***	0.294	0.268
80–84	0.892***	0.565***	1.672***	−0.892***	0.096	1.364*	−1.004***	0.237	0.490*
85–89	1.266***	0.906***	2.183***	−0.764**	0.474*	1.819**	−1.225***	0.082	0.751***
90 and above	1.497***	1.160***	3.242***	−1.237***	0.444+	2.155**	−2.105***	−0.188	1.106***
Female	0.452***	−0.039	−0.715***	−0.019	0.123	−0.431*	−0.260**	0.196+	−0.553***
Whites (ref.)
Blacks	0.279*	0.274**	0.172	−0.329*	0.119	−0.081	−0.466***	−0.059	−0.060
Others	0.865**	0.217	−0.336	0.117	0.126	−1.557*	−0.003	0.705**	−0.274
Hispanics	0.647***	0.312+	−0.600	−0.340	0.036	−0.551+	−0.453*	0.195	−0.395**
Married (ref.)
Separated/divorced	0.456***	0.159	0.596*	−0.238	0.070	0.176	−0.195	0.311*	0.097
Windowed	0.403***	0.198*	0.559**	−0.307*	−0.009	0.047	−0.129	0.165	0.176*
Never married	0.628*	0.076	0.736+	−0.781*	−0.326	0.013	−0.284	0.564*	−0.497*
Proxy interview	−0.340	0.260	1.122*	0.626	1.851***	1.244*	−1.746***	−1.627***	1.173***
Community	−0.197	−0.592***	−0.356	0.628*	−0.089	0.005	0.666***	0.497**	−0.169+
2011 (ref.)
2012	−0.066	0.054	0.391 +	−0.034	0.166	0.019	−0.159	−0.061	0.228*
2013	−0.064	0.018	0.234	0.059	0.305*	0.473+	−0.221*	−0.136	0.067
2014	−0.255+	0.094	0.288	−0.249	−0.230	−0.317	−0.292**	−0.172	−0.306*
2015	−0.202	0.004	0.123	−0.295	−0.230	0.202	−0.250*	−0.485***	−0.109
2016	−0.198	0.204+	0.254	−0.168	0.090	0.501+	−0.170	−0.275+	0.088
2017	0.023	0.011	−0.511	−0.595**	0.322+	−0.139	−0.361**	−0.277+	0.019
2018	−0.246	−0.078	0.521+	0.013	0.437*	0.420	−0.302*	−0.245	0.044
2019	−0.040	0.136	−0.528	−0.053	0.161	−0.326	−0.563***	−0.117	0.092
Constant	−2.909***	−1.525***	−4.601***	0.138	−0.474	−2.812***	−0.847**	−2.574***	−2.273***

^eWithout refers to without functional limitation or disability and FL refers to functional limitation.

⁺p<0.1,

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Model 1, which focuses on those without functional limitation or disability at baseline, indicates that both pain and education matter with respect to follow-up status. Pain increases the chances of functional limitation and/or disability relative to remaining without functional limitation or disability. Higher education decreases the chances of the less favorable transitions, with complete college being a particularly strong protector. Model 2, which considers those with functional limitation, includes the significant interactions between pain and education and therefore the effect of pain on transitions must be viewed within the context of education. For those with the lowest level of education, pain significantly reduces the probability of a transition from functional limitation to without functional limitation or disability. Therefore, for those with the lowest level of education, pain reduces the chances of recovery and therefore being free of pain results in a higher probability of this favorable transition. But, the interactions, which are positive, suggest that the disadvantage for those with pain disappear when someone has higher education. Transitions from functional limitation to disability or to death are not significantly influenced by pain in Model 2. Model 3, focusing on those with disability at baseline, also includes significant interaction effects. Among those with the lowest level of education, the chance of a favorable transition from disability to either functional limitation only or without functional limitation or disability is significantly reduced for those with pain at baseline and therefore recovery chances are higher without pain. Education again dampens this association. A positive and significant interaction effect suggests that pain has less impact on the chances of improving to functional limitation only among those with complete high school but no college degree. Lastly, contrary to the unadjusted findings, these models do not show a significant association between pain and the chances of dying.

To more easily interpret these results, Table 4 transforms the log odds ratios from Table 3 into predicted percentages of follow-up status by baseline status, pain, and education, and presents these in the form of percent distributions. 95% confidence intervals of these percentages are shown in supplemental Table S2. For those without functional limitation or disability at baseline, the advantage of being pain free is obvious since the predicted percentage of remaining this status at follow-up is higher for those without pain than with across educational categories, and these differences are statistically significant at the 95% level. For instance, for those with less than high school education, 66.5% remain without functional limitation or disability if they have pain, while 75.8% remain in this state if they have no pain. The ratio of these percentages is also provided. Again, a ratio less than 1.00 means the chances are lower for those in pain, while greater than 1.00 means they are higher, and the value of the ratio allows us to assess how the magnitude of the pain effect changes with increasing levels of education. For instance, the ratio for those without functional limitation or disability at baseline remaining in that status at follow-up is 0.88, 0.90 and 0.91 across education levels. These ratios are not just less than 1.00 for each education category, they are also fairly flat, which is a function of a lack of interaction between pain status and education for those without functional limitation or disability at baseline, meaning that the disadvantage of having pain is similar across education levels. For those without functional limitation or disability at baseline, the effects of pain are larger with respect to developing functional limitation and/or disability. For instance, the chances of disability at follow-up for those with the lowest level of education is 20.5% among those with pain and 14.2% among those without, for a ratio of 1.44. The difference is statistically significant according to Table S2.

Table 4.

Predicted percentages of follow-up status by baseline status, pain and education^f

Baseline status	Follow-up status	Less than high school			Complete high school but no college degree			College degree
		With pain	Without pain	Ratiog	With pain	Without pain	Ratio	With pain	Without pain	Ratio
Without	Without (%)	66.5	75.8	0.88	72.3	80.4	0.90	77.4	84.6	0.91
	FL only (%)	11.4	8.3	1.37	8.5	6.0	1.40	4.1	2.9	1.43
	Disability (%)	20.5	14.2	1.44	17.8	12.1	1.48	17.9	11.9	1.51
	Died (%)	1.6	1.7	0.92	1.4	1.5	0.94	0.6	0.6	0.96
	Total (%)	100	100		100	100		100	100
FL only	Without (%)	15.8	28.1	0.56	18.6	21.4	0.87	24.7	26.0	0.95
	FL only (%)	39.0	34.1	1.15	40.5	41.2	0.98	31.8	37.5	0.85
	Disability (%)	40.4	31.5	1.28	37.6	32.3	1.16	40.2	32.1	1.25
	Died (%)	4.8	6.3	0.76	3.3	5.1	0.64	3.3	4.4	0.74
	Total (%)	100	100		100	100		100	100
Disability	Without (%)	6.7	11.3	0.59	8.5	12.9	0.66	13.9	18.2	0.77
	FL only (%)	6.1	8.8	0.69	8.8	8.1	1.08	6.5	7.9	0.82
	Disability (%)	79.0	71.9	1.10	74.8	70.9	1.06	73.9	67.3	1.10
	Died (%)	8.2	8.0	1.04	7.9	8.1	0.98	5.7	6.6	0.86
	Total (%)	100	100		100	100		100	100

^fWithout refers to without functional limitation or disability and FL refers to functional limitation.

^gThe ratio of percentages being in a state for those with pain relative to those without.

The effect of the interaction between pain and education can be seen in the percentages shown both for those with functional limitation only and those with disability at baseline. Here, higher education increases the chances of recovery among those with pain. The chances of recovering from functional limitation only to without functional limitation or disability among those with pain at baseline is 15.8%, 18.6%, and 24.7%, and the ratios across education categories increase from 0.56 to 0.87 and 0.95. Supplemental Table S2 also shows that the difference in percentages of this recovery between people with and without pain is significant for those who have less than high school education, but becomes nonsignificant for the two higher educated groups. Thus, as was pointed out with the multinomial results above, the disadvantage of pain decreases as education increases. We also find similar results among those with disability at baseline. The chances of improving to without functional limitation or disability among those with pain at baseline is 6.7%, 8.5% and 13.9% across education categories and the ratios across education categories for this group are 0.59, 0.66 and 0.77. Therefore, pain has a greater impact on reducing the chances of complete recovery for those with lower education, while those with higher education and pain have a higher chance of improving status.

4. Discussion

This paper aims to understand the role of pain in the disablement process, including how pain is implicated in the onset of and recovery from disability, as well as how education modifies these processes. Theoretically, we adopted the conceptualization of disablement from Verbrugge and Jette.⁴⁶ In their model, pain is classified as an impairment that results in the restriction of physical movements, or functional limitation, which in turn influences the ability to perform tasks necessary for daily survival, or disability, the latter of which are influenced by a social and physical environment. We focus on the population aged 65 years and over, which are ages whereby the prevalence of pain and disability and the likelihood of experiencing pain and disability incidence are higher than those younger.^{4, 7, 33} We examine the transition from three baseline states - without functional limitation or disability, with functional limitation only, and with disability - to four different follow-up states - without functional limitation or disability, with functional limitation only, with disability, and deceased.

Our results generally support hypotheses H1 and H2 that suggest pain may forecast a transition to disability and may disadvantage the possibility of recovery for those with disability. This conclusion is reached since net of other covariates, pain significantly increases the likelihood of having a functional limitation or disability at follow-up and decreases the likelihood of recovering from functional limitation or disability. This finding is not surprising and echoes a number of previous studies.^{7, 36} Consistent with earlier literature,^{6, 21, 28, 31, 35} we also found higher education to be protective. Those with more education are less likely to develop functional limitation and disability and more likely to recover from these states compared to their lower educated counterparts.

H3 and H4 considered the moderating impact of education. H3 is not supported. Education does not protect people with pain from transitioning from a state without functional limitation or disability to a less favorable state. This contradicts some previous literature that showed socioeconomic status is a strong predictor of disability among those who have pain.^{19, 29} Nonetheless, there are other studies that suggest education is less predictive in the development of disability than other socioeconomic characteristics such as income or wealth.^{18, 44, 50} Evidence from randomized controlled trial research also shows that education alone may not protect people in the progress from pain to disability.¹⁷ Therefore, there may be differences in effects across different socioeconomic characteristics. But there is clearly some conflicting evidence to this regard in past literature. Our findings are generally supportive of the notion that once an individual is bothered by pain, the step to disablement is highly inevitable and thus educational disparities may be relatively uninfluential.²⁶ More research is needed to disentangle the causal mechanisms leading from pain to disability across education and other socioeconomic indicators.

We do however find support for H4, that education buffers the negative effect of pain on the recovery process among those with functional limitation or disability at baseline. Higher education increases the chances that those with pain and functional limitation recover to without functional limitation or disability, and that those with pain and disability become disability free. Although education does not protect from disability incidence for those with pain, assisting in recovery is suggestive of different mechanisms affecting disablement versus recovery and different coping mechanisms and strategies existing across educational categories once there is an incidence of functional limitation or disability. This is consistent with previous research that indicated education derives some advantages with respect to coping with unfavorable health status.^{6, 28} In face of pain and functional limitation, those with higher education may have advantages over their lower educated counterparts in developing coping strategies such as actively seeking quality medical care. It is also possible that healthcare providers are less sensitive to pain issues reported by those with lower socioeconomic, and thus these people may not obtain the same treatment and care compared to their more advantaged counterparts.⁴²

For people who have already developed disability, higher education may associate with finding and utilizing adjustments to disability in the face of pain. Recall that disability is defined using the Katz ADL conceptualization²⁴ as being able to conduct daily activities like dressing and bathing, which in the disablement process means that disability involves both physical ability and the needs of the environment in which one lives. It is possible that two individuals with the same physical limitations will have different disability outcomes due to differences in how they cope with those limitations. Those with pain that are better educated may be better able to utilize resources that facilitate completing these types of tasks in the face of pain, for instance, by using assistive devices, paying for help, or adopting strategies for rehabilitation.³

There are other findings in our study that were unexpected. We find little impact of pain on the probability of dying in the adjusted models, which is contrary to other studies that suggest a causal association between pain and mortality.^{2, 38} There are several possible explanations. Among those without pain their functional limitations and disability are possibly more so a function of chronic diseases that are more likely to be fatal, such as cancer.^{11, 43} For those with pain, their disablement may be newly developed and easy to recover from.²² Pain might lead one to seek medical treatment more urgently, resulting in better and faster treatment for health problems. The net result could be little difference in the probability of dying by pain status.

While the Verbrugge and Jette⁴⁶ conceptualization of the disablement process served us well in fashioning this analysis, our study also points to several possible disadvantages of this model. First, it is clear that pain and disability are more intertwined than is suggested by the model. While we find that those with pain are more likely to have a disability incidence by follow-up, we also find in supplementary analysis (available upon request) that pain and disability tend to show up in the same survey wave, indicating the likelihood that the process from pain to disability can occur without any or with an extremely short time lag. Second, the model pays little attention to recovery from disablement and the process by which that occurs. Our findings suggest that improvement is common and different from the disablement process, thus warranting separate discussions. There is indeed evidence in the literature that though often short-lasting, recovery from disability is prevalent among older adults, especially among those who are newly disabled, but the odds of recovery decrease with severity of impairment.^{8, 22, 27} The determinants of this improvement process are different from the determinants of disablement.⁸ Given these disadvantages, other frameworks may be considered. One alternative is the International Classification of Functioning, Disability, and Health (or ICF), provided by the World Health Organization,⁴¹ which sees disability as an umbrella term containing health conditions, body functions and structures, activities, and participation. It incorporates all processes happening around disability and all individuals regardless of their stage of functioning.¹³ Therefore, the ICF framework may be better equipped to consider recovering from disability and the intertwined and recurrent nature of pain, disability, and recovery.

The study is subject to limitations. First, NHATS does not measure severity of pain and thus this cannot be incorporated into the analysis. Based on only a binary measure, the relationship between pain and functional limitation/disability may be underestimated. In addition, there is little information about the persistency and cause of pain, thus our measure of pain is a general report that does not show from where disability is derived. Nevertheless, the results broadly reflect the linkage between pain and disability outcomes. Second, our analysis is limited to those who are 65+, thus the findings cannot be generalized to other ages. Given that the prevalence of pain is also rising for the middle-aged population,⁴⁷ it would be important to test the hypotheses in other available data. Third, our analysis is based on one-year transitions and thus does not capture the full dynamics of the disablement and recovery process. Since previous studies suggest that disability may be a recurrent rather than an enduring condition, and recovery is often short-lasting,²² more research is warranted to further examine the fluctuation of disability outcomes through life course and trajectory perspectives.

The study fills a gap in the literature that under-appreciates pain as a critical aspect of the disablement and recovery process and enhances our understanding in the causal connection between pain and disablement.^{48, 51} Our results highlight the important role of pain in the disablement process as well as in recovering from functional limitation or disability. Those with pain are more likely to move from being free of functional limitation or disability to having functional limitations and/or disability, while lack of pain facilitates recovery. However, pain does not necessarily link with higher mortality once adjusting for covariates. Education does not alter the progress of developing disability for those who have pain, but it helps facilitate recovery among those with pain at baseline.

Highlights.

• Pain predicts the development of functional limitation and disability after one year.

• Pain decreases the chances of recovery from functional limitation or disability.

• Education cannot prevent the onset of functional limitation and disability for those with pain.

• Education facilitates the recovery from functional limitation or disability for those with pain.

Perspective.

This article is among the first examining how pain is placed in the disablement process by affecting onset of and recovery from disability. Both paths are affected by pain, but education moderates the association only with respect to the recovery process.