Descriptive statistics
Table 1 below exhibits socio-demographic characteristics of the decedents. Age at death ranged between 54 and 121 (M = 94.33, SD = 7.95). Women comprised 58.75% of the sample. On average, the participants received 1.48 years of schooling (SD = 2.91). Towards end-of-life, over half (58.68%) lived in rural regions while the rest in cities or towns. Most subjects (87.03%) were community-dwelling, and about 90% were primarily cared for by their family. Additionally, as far as the available data have implied, nine out of ten responses were from a next of kin. Given that information on end-of-life status tends to be accurate when provided by family members [7, 28], the proxy responses from the CLHLS were considered reliable.
Table 1
Socio-demographic characteristics of subjects
Characteristics | N | % | Characteristics | N | % |
Age at death (years) | | Relationship between the respondent and the deceased* |
Mean (SD) | 94.35 (9.03) | Relative | 1,106 | 90.07 |
Minimum–maximum | 54–121 | Non-relative | 122 | 9.93 |
Sex | | | Months between the last interview and death |
Male | 4,033 | 41.25 | Mean (SD) | 18.33 (11.29) |
Female | 5,745 | 58.75 | Minimum–maximum | 0-61.70 |
Education (years) | Primary caregiver |
Mean (SD) | 1.48 (2.91) | No caregiver needed | 233 | 2.42 |
Minimum–maximum | 0–25 | Lack of a caregiver | 66 | .69 |
Residence location prior to death | Spouse | 751 | 7.80 |
Rural areas | 5,738 | 58.68 | Children or grandchildren | 7,899 | 82.08 |
Urban areas (cities/towns) | 4,040 | 41.32 | Others from family, or friends | 239 | 2.48 |
Marital status prior to death | Others | 435 | 4.52 |
Unmarried | 7,719 | 80.11 | Place of death | | |
Married | 1,916 | 19.89 | Home | 8,544 | 88.74 |
Living arrangement prior to death | Hospital | 812 | 8.43 |
Alone | 1,245 | 12.97 | Nursing homes | 215 | 2.23 |
With others | 8,352 | 87.03 | Others | 57 | .59 |
Note. N = 9,778. * Information on respondents’ relationship to the deceased was only available for the 2018 wave. |
Table 2 here describes health condition, six indicators of physical functioning, and two items representing quality of death among participants. Over half of the total sample were morbid (57.43%). Difficulties with performing ADLs were prevalent (between 65.62% and 79.64%) in five out of six activities except for continence, with which slightly less than half (48.10%) had trouble. Almost two fifths of the deaths were of less-than-satisfactory qualities, featured by feelings of pain (38.59%) and unconsciousness (36.09%).
Table 2
Descriptive statistics on health condition, physical functioning, and quality of death by sex and location
| Sex | Residence location |
| Male N (%) | Female N (%) | Rural areas N (%) | Urban areas N (%) |
Health condition (N in total sample; %) | |
No disease (4,026; 42.58) | 1,455 (37.38) | 2,571 (46.21) | 2,523 (45.22) | 1,503 (38.77) |
Single disease (2,932; 31.01) | 1,282 (32.94) | 1,650 (29.65) | 1,713 (30.70) | 1,219 (31.44) |
Two or more (2,498; 26.42) | 1,155 (29.68) | 1,343 (24.14) | 1,343 (24.07) | 1,155 (29.79) |
Physical functioning |
Bathing/Showering (N in total sample; %) |
Independent (1,944; 20.36) | 977 (24.89) | 967 (17.21) | 1,204 (21.36) | 740 (18.93) |
Dependent (7,602; 79.64) | 2,949 (75.11) | 4,653 (82.79) | 4,433 (78.64) | 3,169 (81.07) |
Getting dressed and undressed (N in total sample; %) |
Independent (2,615; 27.39) | 1,227 (31.20) | 1,388 (24.72) | 1,546 (27.45) | 1,069 (27.31) |
Dependent (6,932; 72.61) | 2,706 (68.80) | 4,226 (75.28) | 4,087 (72.55) | 2,845 (72.69) |
Getting to the toilet and cleaning (N in total sample; %) |
Independent (2,469; 25.77) | 1,176 (29.82) | 1,293 (22.94) | 1,457 (25.76) | 1,012 (25.80) |
Dependent (7,111; 74.23) | 2,768 (70.18) | 4,343 (77.06) | 4,200 (74.24) | 2,911 (74.20) |
Indoor mobility (N in total sample; %) |
Independent (2,395; 25.06) | 1145 (29.13) | 1,250 (22.21) | 1,434 (25.39) | 961 (24.58) |
Dependent (7,162; 74.94) | 2785 (70.87) | 4,377 (77.79) | 4,213 (74.61) | 2,949 (75.42) |
Controlling bladder/bowel functions (N in total sample; %) |
Independent (4,966; 51.90) | 2,102 (53.45) | 2,864 (50.83) | 3,019 (53.44) | 1,947 (49.68) |
Dependent (4,602; 48.10) | 1,831 (46.55) | 2,771 (49.17) | 2,630 (46.56) | 1,972 (50.32) |
Feeding oneself (N in total sample; %) |
Independent (3,284; 34.38) | 1,483 (37.74) | 1,801 (32.03) | 1,949 (34.54) | 1,335 (34.15) |
Dependent (6,268; 65.62) | 2,446 (62.26) | 3,822 (67.97) | 3,694 (65.46) | 2,574 (65.85) |
Quality of death | |
Freedom from pain (N in total sample; %) | | | |
Yes (5,475; 61.41) | 2,087 (56.94) | 3,388 (64.52) | 3,121 (59.39) | 2,354 (64.30) |
No (3,441; 38.59) | 1,578 (43.06) | 1,863 (35.48) | 2,134 (40.61) | 1,307 (35.70) |
Being conscious (N in total sample; %) | |
Yes (6,085; 63.91) | 2,624 (66.96) | 3,461 (61.78) | 3,646 (64.88) | 2,439 (62.52) |
No (3,436; 36.09) | 1,295 (33.04) | 2,141 (38.22) | 1,974 (35.12) | 1,462 (37.48) |
Note. Percentages (%) are proportions of non-missing values. |
Sex differences emerged towards end-of-life. Women were relatively heathier, as the proportion of women whose close contacts recalled having no disease (46.21%) was larger than that of men (37.38%), while a higher percentage of men were believed to be morbid. On the other hand, men functioned much better, because a higher fraction of men than women could perform all ADLs without assistance: the sex gap regarding proportions varied between 2.62% and 7.68%. Discrepancies persisted till death. Painlessness was more often witnessed among female decedents (64.52% of women but 56.94% of men), while consciousness among male ones (66.96% vs. 61.78%).
Compared to sex disparities, residence-based divergences were not as pronounced. Urban dwellers were more often morbid: within three months prior to death, 31.44% of city-settlers lived with one disease and 29.79% with two or more, while the respective figures were 30.70% and 24.07% for rural residents. Circumstances with physical functioning (indicated by six ADL measures) were less straightforward. Higher percentages of older people in urban areas could perform three ADLs independently than their peers in rural locations (bathing/showering, indoor mobility, and continence), and the percentage differences ranged from .81–3.76%. By comparison, urban-rural divides in the remaining three ADLs (dressing, getting to the toilet, and eating) were minimal—percentage differences at .39% and lower. When dying, a larger share of urban decedents was painless (64.30% vs. 59.39%), but proportions of consciousness at death were substantially similar in urban and rural regions (62.52% in cities or towns and 64.88% in villages).
Cfa And Mimic-model Method
CFA was carried out to determine whether a common factor, physical functioning, underlay the six items of ADLs (see the graphic representation in Figure 1, Panel A). All fit indices were optimal (CFI= 1.000; TLI= 1.000; RMSEA= .038, 90% CI: [.033, .044]; SRMR = .010), apart from the statistically significant chi-square statistics (χ2 [df] = 135.319 [9], p < .001). Because model chi-square tests the exact-fit hypothesis, which is sensitive to sample sizes larger than 300, the specified model was still recognized as a reasonably good fit for the data [29].
The MIMIC-method was then applied to identify the effect of covariates (sex and residence) at the observed level (i.e., in one or more indicators of physical functioning; see Figure 1, Panel B for an example) while holding constant the latent factor. The results were summed in Table 3 below (values of other path coefficients are substantially similar with the original CFA model and are thus omitted to save space but are available on request). The following trends emerged: proxy respondents of older women were more inclined to recall problems with bathing/showering and indoor mobility, but surrogates of older men were more liable to report issues with continence and feeding. In terms of location, surrogate respondents for rural settlers noticed more difficulties with dressing and getting to the toilet, while those for urban residents suggested more trouble with bathing/showering and continence. These four paths were retained in the subsequent SEM analysis.
Table 3
Fit indices and slope coefficients for MIMIC models testing sex-related or residence-based response differences on each indicator of physical functioning
Indicator | χ2SB | df | RMSEA [90% CI] | CFI | TLI | b | β | p |
Sex | | | | | | | | | |
Bathing/Showering | 170.292*** | 13 | .036 [.031, .040] | 1.000 | .999 | - .096*** | - .047*** | < .001 |
Dressing | 189.440*** | 13 | .038 [.033, .042] | 1.000 | .999 | - .002 | - .001 | .889 |
Getting to the toilet | 187.955*** | 13 | .037 [.033, .042] | 1.000 | .999 | - .023 | - .011 | .065 |
Mobility | 187.677*** | 13 | .037 [.033, .042] | 1.000 | .999 | - .029* | - .014* | .031 |
Continence | 170.974*** | 13 | .036 [.031, .040] | 1.000 | .999 | .106*** | .052*** | < .001 |
Feeding | 188.831*** | 13 | .038 [.033, .042] | 1.000 | .999 | .034* | .017* | .046 |
Residence location | | | | | | | | | |
Bathing/Showering | 161.569*** | 13 | .035 [.030, .039] | 1.000 | .999 | - .060** | - .029** | .002 |
Dressing | 162.569*** | 13 | .035 [.030, .039] | 1.000 | .999 | .040** | .019** | .004 |
Getting to the toilet | 161.153*** | 13 | .034 [.030, .039] | 1.000 | .999 | .046*** | .023*** | < .001 |
Mobility | 166.752*** | 13 | .035 [.030, .040] | 1.000 | .999 | .012 | .006 | .384 |
Continence | 165.912*** | 13 | .035 [.030, .040] | 1.000 | .999 | - .075** | - .037** | .001 |
Feeding | 166.747*** | 13 | .035 [.030, .040] | 1.000 | .999 | .029 | .014 | .090 |
Note. χ2SB, model chi-square for Weighted Least Square Mean Variance (WLSMV) estimation; CI, confidence interval; b, unstandardized estimates; β, standardized estimates. ** p < .01, *** p < .001. |
SEM
SEM was performed to analyse the associations between life prior to death and quality of death. As depicted in Figure 2, the model fitted the data well, χ2 (df) = 293.502 (26), p < .001; CFI= .999; TLI= .999; RMSEA= .033, 90% CI: [.029, .036]; SRMR = .015. Both unstandardized estimates (b) and standardized estimates (β) for the hypothesized paths are displayed in Table 4, and Table 5 (p.18) summarises the direct, indirect, and total effects between variables/factors in interest.
Table 4
Parameter estimates for the associations between life prior to death and quality of death
Parameter | b | S.E.b | β | S.E.β | p |
Path coefficient (regression) | | | |
1. PF → B | 1.000 | .000 | .961*** | .003 | < .001 |
2. PF → D | 1.745*** | .147 | .988*** | .002 | < .001 |
3. PF → G | 1.999*** | .241 | .991*** | .002 | < .001 |
4. PF → M | 1.507*** | .107 | .984*** | .002 | < .001 |
5. PF → C | .385*** | .018 | .820*** | .007 | < .001 |
6. PF → F | .771*** | .039 | .945*** | .003 | < .001 |
7. HC → PF | - .736*** | .061 | - .200*** | .014 | < .001 |
8. HC → FP | - .181*** | .016 | - .177*** | .016 | < .001 |
9. HC → BC | - .097*** | .017 | - .089*** | .015 | < .001 |
10. PF → FP | .023*** | .005 | .082*** | .017 | < .001 |
11. PF → BC | .115*** | .007 | .387*** | .015 | < .001 |
12. Sex → B | - .278*** | .068 | - .035*** | .009 | < .001 |
13. Sex → M | - .077 | .073 | - .007 | .006 | .292 |
14. Sex → C | .172*** | .042 | .049*** | .012 | < .001 |
15. Sex → F | .116* | .055 | .019* | .009 | .036 |
16. Sex → HC | - .196*** | .024 | - .096*** | .012 | < .001 |
17. Sex → PF | - .902*** | .108 | - .120*** | .013 | < .001 |
18. Sex → FP | .187*** | .028 | .089*** | .013 | < .001 |
19. Sex → BC | - .087** | .030 | - .039** | .013 | .003 |
20. Residence location → B | - .261** | .079 | - .033** | .010 | .001 |
21. Residence location → D | .093 | .102 | .007 | .008 | .362 |
22. Residence location → G | .168 | .107 | .011 | .007 | .112 |
23. Residence location → C | - .131** | .040 | - .037** | .011 | .001 |
24. Residence location → HC | .164*** | .024 | .080*** | .012 | < .001 |
25. Residence location → PF | .028 | .098 | .004 | .013 | .778 |
26. Residence location → FP | .169*** | .028 | .081*** | .013 | < .001 |
27. Residence location → BC | - .047 | .029 | - .021 | .013 | .103 |
Path coefficient (covariance) | | | | |
28. FP ↔ BC | .113*** | .019 | .113*** | .019 | < .001 |
Note. S.E., standard error; b, unstandardized estimates; β, standardized estimates; B, bathing/showering; D, dressing; G, getting to the toilet; M, indoor mobility; C, continence; F, feeding; HC, health condition; PF, physical functioning; FP, freedom from pain; BC, being conscious. * p < .05, ** p < .001, *** p < .001. |
Table 5
Direct, indirect, and total effects of life prior to death on quality of death
Effect | b | S.E.b | β | S.E.β | p |
HC → FP | | | | | |
1. Total | - .198*** | .016 | - .194*** | .015 | < .001 |
2. Direct | - .181*** | .016 | - .177*** | .016 | < .001 |
3. Indirect (HC → PF → FP) | - .017*** | .004 | - .016*** | .004 | < .001 |
HC → BC | | | | | |
4. Total | - .181*** | .017 | - .166*** | .015 | < .001 |
5. Direct | - .097*** | .017 | - .089*** | .015 | < .001 |
6. Indirect (HC → PF → BC) | - .085*** | .007 | - .077*** | .006 | < .001 |
Sex → PF | | | | | |
7. Total | - .757*** | .106 | - .101*** | .013 | < .001 |
8. Direct | - .902*** | .108 | - .120*** | .013 | < .001 |
9. Indirect (sex → HC → PF) | .145*** | .021 | .019*** | .003 | < .001 |
Sex → FP | | | | | |
10. Total | .205*** | .028 | .098*** | .013 | < .001 |
11. Direct | .187*** | .028 | .089*** | .013 | < .001 |
12. Indirect (total) | .018* | .007 | .009* | .004 | .013 |
13. Indirect (sex → HC → FP) | .036*** | .005 | .017*** | .003 | < .001 |
14. Indirect (sex → PF → FP) | - .021*** | .005 | - .010*** | .002 | < .001 |
15. Indirect (sex → HC → PF → FP) | .003*** | .001 | .002*** | .000 | < .001 |
Sex → BC | | | | | |
16. Total | - .155*** | .030 | - .069*** | .013 | < .001 |
17. Direct | - .087** | .030 | - .039** | .013 | .003 |
18. Indirect (total) | - .068*** | .013 | - .030*** | .006 | < .001 |
19. Indirect (sex → HC → BC) | .019*** | .004 | .008*** | .002 | < .001 |
20. Indirect (sex → PF → BC) | - .104*** | .012 | - .046*** | .005 | < .001 |
21. Indirect (sex → HC → PF → BC) | .017*** | .002 | .007*** | .001 | < .001 |
Residence location → PF | | | | | |
22. Total | - .093 | .098 | - .012 | .013 | .343 |
23. Direct | .028 | .098 | - .004 | .013 | .778 |
24. Indirect (residence location → HC → PF) | - .121*** | .020 | - .016*** | .003 | < .001 |
Residence location → FP | | | | | |
25. Total | .137*** | .028 | .066*** | .013 | < .001 |
26. Direct | .169*** | .028 | .081*** | .013 | < .001 |
27. Indirect (total) | - .032*** | .006 | - .015*** | .003 | < .001 |
28. Indirect (residence location → HC → FP) | - .030*** | .005 | - .014*** | .002 | < .001 |
29. Indirect (residence location → PF → FP) | - .001 | .002 | .000 | .001 | .778 |
30. Indirect (residence location → HC → PF → FP) | - .003*** | .001 | - .001*** | .000 | < .001 |
Residence location → BC | | | | | |
31. Total | - .074* | .029 | - .033* | .013 | .012 |
32. Direct | - .047 | .029 | - .021 | .013 | .103 |
33. Indirect (total) | - .027* | .012 | - .012* | .005 | .028 |
34. Indirect (residence location → HC → BC) | - .016*** | .004 | - .007*** | .002 | < .001 |
35. Indirect (residence location → PF → BC) | .003 | .011 | .001 | .005 | .778 |
36. Indirect (residence location → HC → PF → BC) | - .014*** | .002 | - .006*** | .001 | < .001 |
Note. S.E., standard error; b, unstandardized estimates; β, standardized estimates; B, bathing/showering; D, dressing; G, getting to the toilet; M, indoor mobility; C, continence; F, feeding; HC, health condition; PF, physical functioning; FP, freedom from pain; BC, being conscious. * p < .05, ** p < .001, *** p < .001. |
The results indicated that different dimensions of life prior to death predicted death qualities (i.e., painlessness and consciousness) to varying degrees. Freedom from diseases contributed to pain-free deaths to a moderately larger extent than to dying with consciousness (|β|s = .194 and .166, both ps < .001), but enhanced functioning capacity was associated more intensely with consciousness maintain rather than pain reduction (|β|s = .387 and .082, both ps < .001). Notably, the abovementioned, morbidity’s effects on both quality markers of death were mediated by functional limitations: morbid individuals’ tendency towards painful deaths was aggravated by problems they had with ADLs (|β| = .016, p < .001), and the amplitude of negative mediation of consciousness loss was even stronger at end-of-life (|β| = .077, p < .001). Nonetheless, both mediatory effects were smaller compared to the direct association between morbidity and a less-than-satisfactory death: the indirect effect of health condition on pain through physical functioning was minimal (|β| = .016, p < .001) compared to the direct effect (|β| = .177, p < .001), although the respective indirect effect on consciousness was only slightly weaker than the direct one (.077 vs .089, both |β|s were statistically significant at p = .001).
Older men and women took divergent trajectories of dying. Sex influenced physical functioning directly and indirectly through health condition: reported incidence of chronic diseases was slightly lower among older women than men (|β| = .096, p < .001), and this modest health advantage only improved women’s physical functioning trivially (|β| = .019, p < .001). Given men’s initial relatively pronounced advantage in functioning (|β| = .120, p < .001), women still functioned worse than men (|β| = .101, p < .001). These sex disparities indirectly impacted the quality of death: older women were less painful than men at death (|β| = .098, p < .001); yet men more often remained conscious (|β| = .069, p < .001).
Compared to sex, residence location played a much lesser role during end-of-life. The rural dwellers appeared healthier (|β| = .080, p < .001), which indirectly enhanced their functional capacities (|β| = .016, p < .001). However, location did not influence physical function directly (|β| = .004, p = .778). On death quality, location had both direct and indirect effects regarding pain: rural residents were more prone to painful deaths than urban settlers (|β| = .081, p < .001), but the former’s better health marginally alleviated that effect to |β| = .066 (p < .001) because freedom from diseases marginally enhanced physical functioning, thereby relieving pain. Location did not directly affect consciousness when dying (|β| = .021, p = .103), but given its impact on morbidity, as well as morbidity’s effect on functional limitations (|β| = .012, p = .028), deaths in villages were slightly more associated with consciousness (|β| = .033, p = .012).