Introduction
Methods
Study design and estimation sample
Defining outcomes: source and target measures
Statistical analysis: exploratory data analysis—conceptual overlap
DLQI | Utility score | |||||
---|---|---|---|---|---|---|
EQ-5D-5L | Mobility | Self-care | Usual activities | Pain/discomfort | Anxiety/depression | |
Total score | − 0.514 | 0.362 | 0.405 | 0.444 | 0.414 | 0.392 |
symptoms | − 0.337 | 0.255 | 0.237 | 0.311 | 0.326 | 0.213 |
Feelings | − 0.343 | 0.193 | 0.245 | 0.255 | 0.246 | 0.336 |
daily activities | − 0.364 | 0.311 | 0.337 | 0.343 | − 0.266 | 0.248 |
Clothing | − 0.295 | 0.214 | 0.248 | 0.269 | − 0.241 | 0.212 |
Social activities | − 0.441 | 0.339 | 0.346 | 0.382 | 0.325 | 0.330 |
Sport | − 0.310 | 0.241 | 0.258 | 0.328 | 0.272 | 0.171 |
Work and school | − 0.382 | 0.320 | 0.310 | 0.380 | 0.320 | 0.223 |
Personal relationship | − 0.323 | 0.255 | 0.292 | 0.310 | 0.225 | 0.226 |
Sex | − 0.290 | 0.251 | 0.229 | 0.278 | 0.241 | 0.162 |
Treatment | − 0.331 | 0.260 | 0.299 | 0.311 | 0.277 | 0.249 |
Modeling approaches and performance
Type of model | Prediction | Modeling distribution |
---|---|---|
OLS | Utility | Normal |
Tobit | Utility | Normal |
Two-part: GLM (1)–OLS (2) | Disutility | Part 1: binomial, part 2: normal |
Two-part: GLM (1)–OLS (2) | Disutility | Part 1: binomial, part 2: lognormal |
Two-part: GLM (1)–GLM (2) | Disutility | Part 1: binomial, part 2: gamma |
Regression mixture | Utility | Normal |
Level | Available data | Independent variables |
---|---|---|
Level 1 | Total DLQI | Total DLQI |
Level 2 | Total DLQI, age, and sex | Total DLQI, age, and sex |
Level 3 | DLQI items | DLQI1, DLQI 2, DLQI3, DLQI4, DLQI5, DLQI6, DLQI7, DLQI8, DLQI9, DLQI10 |
Level 4 | DLQI items, age, and sex | DLQI1, DLQI 2, DLQI3, DLQI4, DLQI5, DLQI6, DLQI7, DLQI8, DLQI9, DLQI10, age, sex |
Results
Descriptive statistics: characterizing the cohort
Descriptive statistics | |
---|---|
Characteristics | Estimation sample (out of 9759) |
Respondents (n) | 1232 |
Age (mean ± SD) | 48.28 ± 14.98 |
Women (proportion) | 67.37% |
EQ-5D-5L (mean ± SD) | 0.78 ± 021 |
Median | 0.83 |
Kurtosis | 3.55 |
Skewness | − 1.73 |
Shapiro Wilk | p value < 2.2e−16 |
DLQI score (mean ± SD) | 7.23 ± 6.19 |
Median | 5 |
Kurtosis | 1.37 |
Skewness | 4.63 |
Shapiro Wilk | p value < 2.2e−16 |
Conceptual overlap
Model selection and performance
Regression type | Mean | SD | EQ-5D-5L | ||||
---|---|---|---|---|---|---|---|
Min | Max | RMSE | MAE | Rank | |||
Observed EQ-5D-5L | 0.777 | 0.2097 | − 0.285 | 1 | |||
Total DLQI | |||||||
OLS | 0.776 | 0.112 | 0.366 | 0.907 | 0.178 | 0.127 | 12 |
Tobit | 0.791 | 0.122 | 0.342 | 0.933 | 0.178 | 0.124 | 11 |
Two-part: GLM (logistic)–OLS (normal) | 0.793 | 0.202 | − 0.029 | 1.424 | 0.266 | 0.198 | 20 |
Two-part: GLM (logistic)–OLS (lognormal) | 0.816 | 0.128 | 0.304 | 1 | 0.230 | 0.158 | 13 |
Two-part: GLM (logistic)–GLM (gamma) | 0.825 | 0.142 | − 0.167 | 1 | 0.253 | 0.180 | 17 |
Regression mixture | 0.794 | 0.158 | 0.108 | 0.933 | 0.113 | 0.080 | 2 |
Total DLQI + age + sex | |||||||
OLS | 0.776 | 0.116 | 0.342 | 0.972 | 0.175 | 0.125 | 10 |
Tobit | 0.791 | 0.127 | 0.313 | 1.010 | 0.176 | 0.123 | 9 |
Two-part: GLM (logistic)–OLS (normal) | 0.793 | 0.202 | 0.046 | 1.313 | 0.266 | 0.198 | 20 |
Two-part: GLM (logistic)–OLS (lognormal) | 0.816 | 0.128 | 0.244 | 1 | 0.230 | 0.158 | 13 |
Two-part: GLM (logistic)–GLM (gamma) | 0.818 | 0.150 | − 0.071 | 1 | 0.257 | 0.186 | 18 |
Regression Mixture | 0.794 | 0.161 | 0.071 | 0.969 | 0.113 | 0.079 | 1 |
DLQI items | |||||||
OLS | 0.776 | 0.130 | 0.233 | 1.012 | 0.165 | 0.118 | 8 |
Tobit | 0.790 | 0.141 | 0.228 | 1.079 | 0.166 | 0.117 | 5 |
Two-part: GLM (logistic)–OLS (normal) | 0.793 | 0.202 | − 0.063 | 1.335 | 0.266 | 0.198 | 20 |
Two-part: GLM (logistic)–OLS (lognormal) | 0.816 | 0.128 | 0.210 | 1 | 0.230 | 0.158 | 13 |
Two-part: GLM (logistic)–GLM (gamma) | 0.818 | 0.160 | − 0.792 | 1 | 0.260 | 0.181 | 18 |
Regression mixture | 0.790 | 0.163 | 0.140 | 0.994 | 0.113 | 0.083 | 4 |
DLQI items + age + sex | |||||||
OLS | 0.776 | 0.131 | 0.220 | 1.003 | 0.164 | 0.118 | 5 |
Tobit | 0.791 | 0.142 | 0.218 | 1.070 | 0.166 | 0.117 | 5 |
Two-part: GLM (logistic)–OLS (normal) | 0.793 | 0.202 | 0.007 | 1.319 | 0.266 | 0.198 | 20 |
Two-part: GLM (logistic)–OLS (lognormal) | 0.816 | 0.128 | 0.182 | 1 | 0.230 | 0.158 | 13 |
Two-part: GLM (logistic)–GLM (gamma) | 0.820 | 0.168 | − 0.809 | 1 | 0.267 | 0.187 | 24 |
Regression mixture | 0.791 | 0.163 | 0.085 | 0.996 | 0.112 | 0.082 | 2 |