Text box 1. Contributions to the literature
|
---|
• Previous studies ignored the association between medical insurance and IADL disability. This study finds that the middle-aged and older adults who were covered by URRBMI had a reduced possibility of suffering from IADL disability when compared with those without URRBMI. |
• The heterogeneity analysis reveals that URRBMI produced a statistically stronger effect on IADL disability for the older adults when compared with the middle-aged adults. |
• The impact channel analysis reveals that physical exercise was a channel involving the effect of URRBMI on IADL disability. |
• From the perspective of methods utilization, this study used the generalized random forest model from machine learning to conduct a robustness check, which encourages the use of more machine learning methods in health issue. |
Background
Methods
Data source
Variables
Dependent variable
Independent variable
Control variables
Variable | VIF | 1/VIF |
---|---|---|
Age | 1.19 | 0.84 |
Education status | 1.17 | 0.86 |
Marital status | 1.15 | 0.87 |
Gender | 1.10 | 0.91 |
Residency area | 1.08 | 0.92 |
Having tap water | 1.04 | 0.96 |
PM2.5 concentrations | 1.04 | 0.96 |
Household income | 1.03 | 0.97 |
URRBMI | 1.03 | 0.97 |
Mean VIF | 1.09 |
Channel variables
Statistical analyses
Results
Descriptive statistics
Variable | Full sample (N = 1,924) | With URRBMI (N = 1,593) | Without URRBMI (N = 331) |
---|---|---|---|
IADL disability | |||
Noa, n (%) | 1,495 (77.70) | 1,267 (79.54) | 228 (68.88) |
Yes, n (%) | 429 (22.30) | 326 (20.46) | 103 (31.12) |
Age | |||
45–64, n (%) | 1,235 (64.19) | 1,033 (64.85) | 202 (61.03) |
≥ 65, n (%) | 689 (35.81) | 560 (35.15) | 129 (38.97) |
Gender | |||
Femalea, n (%) | 1,015 (52.75) | 836 (52.48) | 179 (54.08) |
Male, n (%) | 909 (47.25) | 757 (47.52) | 152 (45.92) |
Marital status | |||
Single, divorced or widoweda, n (%) | 257 (13.36) | 190 (11.93) | 67 (20.24) |
Married, n (%) | 1,667 (86.64) | 1,403 (88.07) | 264 (79.76) |
Residency area | |||
Rural areasa, n (%) | 1,187 (61.69) | 973 (61.08) | 214 (64.65) |
Urban areas, n (%) | 737 (38.31) | 620 (38.92) | 117 (35.35) |
Education status | |||
Literatea, n (%) | 370 (19.23) | 289 (18.14) | 81 (24.47) |
Illiterate, n (%) | 1,554 (80.77) | 1,304 (81.86) | 250 (75.53) |
Household income (RMB) | |||
Mean (SD) | 27,789.56 (46,934.82) | 29,091.85 (49,294.55) | 21,522.03 (32,657.31) |
PM2.5 concentrations (μg/m3) | |||
Mean (SD) | 34.37 (14.65) | 34.89 (14.98) | 31.87 (12.68) |
Having tap water | |||
Noa, n (%) | 334 (17.36) | 271 (17.01) | 63 (19.03) |
Yes, n (%) | 1,590 (82.64) | 1,322 (82.99) | 268 (80.97) |
Health services utilization | |||
Noa, n (%) | 1,428 (74.22) | 1,152 (72.32) | 276 (83.38) |
Yes, n (%) | 496 (25.78) | 441 (27.68) | 55 (16.62) |
Physical exercise | |||
Noa, n (%) | 184 (9.56) | 146 (9.17) | 38 (11.48) |
Yes, n (%) | 1,740 (90.44) | 1,447 (90.83) | 293 (88.52) |
Regression results of the association between URRBMI and IADL disability
Variable | Model 1 | Model 2 | Model 3 | Model 4 |
---|---|---|---|---|
URRBMI | 0.5696** | 0.5583** | 0.6039** | 0.6014** |
(0.0845) | (0.0920) | (0.1008) | (0.1034) | |
Age | 1.0660** | 1.0605** | 1.0631** | |
(0.0076) | (0.0074) | (0.0071) | ||
Gender | 0.4834** | 0.5555** | 0.5557** | |
(0.0555) | (0.0662) | (0.0673) | ||
Marital status | 1.1825 | 1.2423 | 1.2905 | |
(0.1837) | (0.1980) | (0.2097) | ||
Residency area | 0.7456 | 0.7913 | 0.8927 | |
(0.1348) | (0.1460) | (0.1577) | ||
Education status | 0.6136** | 0.6260** | ||
(0.1031) | (0.1044) | |||
Household income | 0.9270** | 0.9320** | ||
(0.0177) | (0.0171) | |||
PM2.5 concentrations | 0.9987 | |||
(0.0069) | ||||
Having tap water | 0.4713** | |||
(0.0903) | ||||
Constant | 0.4518** | 0.0113** | 0.0354** | 0.0507** |
(0.0533) | (0.0057) | (0.0188) | (0.0310) | |
Number of observations | 1,924 | 1,924 | 1,924 | 1,924 |
Wald chi-squared | 14.39** | 117.26** | 152.29** | 208.52** |
Pseudo R-squared | 0.0083 | 0.0789 | 0.0934 | 0.1068 |
IV-Probit analysis
Variable | First stage | Second stage |
---|---|---|
URRBMI | IADL disability | |
(1) | (2) | |
Participation rate of URRBMI at the community level | 0.9863** | |
(0.0210) | ||
URRBMI | -0.3419** | |
(0.1180) | ||
Age | 0.0006 | 0.0355** |
(0.0007) | (0.0039) | |
Gender | 0.0040 | -0.3371** |
(0.0122) | (0.0713) | |
Marital status | 0.0157 | 0.1438 |
(0.0183) | (0.1007) | |
Residency area | -0.0082 | -0.0582 |
(0.0124) | (0.0724) | |
Education status | 0.0342* | -0.2801** |
(0.0159) | (0.0838) | |
Household income | 0.0032 | -0.0407** |
(0.0019) | (0.0109) | |
PM2.5 concentrations | 0.0001 | -0.0009 |
(0.0004) | (0.0023) | |
Having tap water | 0.0055 | -0.4436** |
(0.0157) | (0.0840) | |
Constant | -0.0952 | -1.7003** |
(0.0576) | (0.3176) | |
Number of observations | 1,924 | 1,924 |
F test | 257.08** | |
R-squared | 0.5473 | |
Adj R-squared | 0.5451 | |
Wald chi-squared | 199.00** | |
Wald test of exogeneity | 0.19 |
Robustness checks
Matching method | Sample | Treated | Control | ATT | S.E | T-stat |
---|---|---|---|---|---|---|
K-nearest neighbor matching | Unmatched | 0.2046 | 0.3112 | -0.1065 | 0.0250 | -4.25 |
Matched | 0.2049 | 0.2973 | -0.0924 | 0.0313 | -2.95** | |
Radius matching | Unmatched | 0.2046 | 0.3112 | -0.1065 | 0.0250 | -4.25 |
Matched | 0.2050 | 0.2988 | -0.0938 | 0.0295 | -3.18** | |
Kernel matching | Unmatched | 0.2046 | 0.3112 | -0.1065 | 0.0250 | -4.25 |
Matched | 0.2049 | 0.2909 | -0.0860 | 0.0288 | -2.99** | |
Nearest-neighbor matching within caliper | Unmatched | 0.2046 | 0.3112 | -0.1065 | 0.0250 | -4.25 |
Matched | 0.2050 | 0.2964 | -0.0913 | 0.0313 | -2.92** |
Variable | Model 1 | Model 2 | Model 3 | Model 4 |
---|---|---|---|---|
URRBMI | -0.1111** | -0.1238** | -0.1086** | -0.1102** |
(0.0336) | (0.0358) | (0.0314) | (0.0310) | |
Cluster | No | No | No | Yes |
Number of decision trees | 500 | 1,000 | 2,000 | 2,000 |
Number of observations | 1,924 | 1,924 | 1,924 | 1,924 |
Heterogeneity analysis
Variable | Model 1 | Model 2 | Model 3 | Model 4 |
---|---|---|---|---|
Middle-aged adults | Older adults | Rural areas | Urban areas | |
URRBMI | 0.5690* | 0.5815** | 0.6392* | 0.5412* |
(0.1520) | (0.1205) | (0.1401) | (0.1415) | |
Age | 1.1055** | 1.0535** | 1.0749** | 1.0437** |
(0.0312) | (0.0121) | (0.0092) | (0.0116) | |
Gender | 0.6157* | 0.5220** | 0.5616** | 0.5184** |
(0.1172) | (0.0878) | (0.0904) | (0.0949) | |
Marital status | 1.1689 | 1.2838 | 1.4294 | 1.1440 |
(0.5090) | (0.2324) | (0.2954) | (0.3069) | |
Residency area | 1.2297 | 0.7399 | ||
(0.3084) | (0.1433) | |||
Education status | 0.5812* | 0.6549* | 0.6843* | 0.5024* |
(0.1544) | (0.1207) | (0.1272) | (0.1689) | |
Household income | 0.9439** | 0.9201** | 0.9208** | 0.9491* |
(0.0206) | (0.0257) | (0.0233) | (0.0247) | |
PM2.5 concentrations | 0.9789* | 1.0101 | 0.9993 | 0.9985 |
(0.0090) | (0.0071) | (0.0086) | (0.0110) | |
Having tap water | 0.3490** | 0.5916** | 0.4494** | 0.5205* |
(0.0964) | (0.1156) | (0.1038) | (0.1710) | |
Constant | 0.0139** | 0.0654** | 0.0228** | 0.1683 |
(0.0224) | (0.0619) | (0.0188) | (0.1686) | |
Number of observations | 890 | 1034 | 1187 | 737 |
Wald chi-squared | 61.40** | 81.06** | 139.32** | 85.19** |
Pseudo R-squared | 0.0970 | 0.0817 | 0.1183 | 0.0880 |
Impact channel analysis
Variable | Model 1 | Model 2 | Model 3 |
---|---|---|---|
URRBMI | -0.0898** | -0.1087** | -0.0854** |
(0.0291) | (0.0262) | (0.0291) | |
Age | 0.0099** | 0.0094** | 0.0091** |
(0.0011) | (0.0011) | (0.0011) | |
Gender | -0.0881** | -0.0798** | -0.0874** |
(0.0185) | (0.0185) | (0.0179) | |
Marital status | 0.0337 | 0.0291 | 0.0310 |
(0.0285) | (0.0303) | (0.0280) | |
Residency area | -0.0146 | -0.0187 | -0.0123 |
(0.0260) | (0.0188) | (0.0255) | |
Education status | -0.0958** | -0.1025** | -0.0958** |
(0.0331) | (0.0282) | (0.0326) | |
Household income | -0.0110** | -0.0104** | -0.0101** |
(0.0032) | (0.0029) | (0.0030) | |
PM2.5 concentrations | -0.0002 | 0.0002 | -0.0004 |
(0.0011) | (0.0007) | (0.0010) | |
Having tap water | -0.1304** | -0.1273** | -0.1327** |
(0.0365) | (0.0271) | (0.0356) | |
Health services utilization | 0.1638** | ||
(0.0225) | |||
Physical exercise | -0.1802** | ||
(0.0309) | |||
Constant | 0.0011 | -0.0093 | 0.2112 |
(0.1037) | (0.0938) | (0.1093) | |
1 − β/α | -0.2105 | 0.0490 | |
Number of observations | 1,924 | 1,924 | 1,924 |
F test | 26.88** | 29.97** | 28.40** |
R-squared | 0.1138 | 0.1428 | 0.1296 |