ODMACS Database Analysis: WHODAS Items & IRT
UK/Brazil Team
2025-02-18
Overview
This document presents the analysis of the ODMACS database including: - Data loading and preprocessing - Thematic subsetting of variables - WHODAS 2.0 items analysis with IRT using mirt and EGAnet - Distribution of WHODAS items (frequency tables and stacked bar charts) - Likert analysis of comparison items
All code is folded by default. Click the arrow at the top right of each code chunk to view the code.
1. Data Loading and Preprocessing
2. Thematic Subsets
We create several subsets: basic information, persisting symptoms, resolved symptoms, quality of life, functional status, and dyspnoea.
3. WHODAS Items & IRT Analysis
We focus on the 12 core WHODAS items. First, we convert them to numeric, compute a simple sum score, then fit a unidimensional graded‐response IRT model, extract item parameters, plot item characteristic curves and the test information function, and derive factor scores (theta).
## Iteration: 1, Log-Lik: -13131.556, Max-Change: 4.34382Iteration: 2, Log-Lik: -11983.536, Max-Change: 3.14197Iteration: 3, Log-Lik: -11737.767, Max-Change: 1.42252Iteration: 4, Log-Lik: -11644.629, Max-Change: 1.29672Iteration: 5, Log-Lik: -11600.219, Max-Change: 0.45293Iteration: 6, Log-Lik: -11575.945, Max-Change: 0.55321Iteration: 7, Log-Lik: -11558.949, Max-Change: 0.43284Iteration: 8, Log-Lik: -11546.928, Max-Change: 0.45045Iteration: 9, Log-Lik: -11538.094, Max-Change: 0.31023Iteration: 10, Log-Lik: -11531.523, Max-Change: 0.36852Iteration: 11, Log-Lik: -11526.210, Max-Change: 0.24028Iteration: 12, Log-Lik: -11521.878, Max-Change: 0.27897Iteration: 13, Log-Lik: -11518.290, Max-Change: 0.17247Iteration: 14, Log-Lik: -11515.284, Max-Change: 0.21801Iteration: 15, Log-Lik: -11512.641, Max-Change: 0.14107Iteration: 16, Log-Lik: -11510.503, Max-Change: 0.16987Iteration: 17, Log-Lik: -11508.509, Max-Change: 0.10486Iteration: 18, Log-Lik: -11506.820, Max-Change: 0.13000Iteration: 19, Log-Lik: -11505.465, Max-Change: 0.08994Iteration: 20, Log-Lik: -11504.073, Max-Change: 0.10273Iteration: 21, Log-Lik: -11502.894, Max-Change: 0.07921Iteration: 22, Log-Lik: -11501.779, Max-Change: 0.09256Iteration: 23, Log-Lik: -11500.777, Max-Change: 0.06180Iteration: 24, Log-Lik: -11499.871, Max-Change: 0.06909Iteration: 25, Log-Lik: -11497.909, Max-Change: 0.04381Iteration: 26, Log-Lik: -11497.378, Max-Change: 0.05151Iteration: 27, Log-Lik: -11496.850, Max-Change: 0.04088Iteration: 28, Log-Lik: -11494.504, Max-Change: 0.02416Iteration: 29, Log-Lik: -11494.334, Max-Change: 0.01217Iteration: 30, Log-Lik: -11494.147, Max-Change: 0.01048Iteration: 31, Log-Lik: -11493.885, Max-Change: 0.00940Iteration: 32, Log-Lik: -11493.746, Max-Change: 0.00880Iteration: 33, Log-Lik: -11493.622, Max-Change: 0.00842Iteration: 34, Log-Lik: -11493.048, Max-Change: 0.00690Iteration: 35, Log-Lik: -11492.991, Max-Change: 0.00626Iteration: 36, Log-Lik: -11492.943, Max-Change: 0.00553Iteration: 37, Log-Lik: -11492.898, Max-Change: 0.00494Iteration: 38, Log-Lik: -11492.862, Max-Change: 0.00584Iteration: 39, Log-Lik: -11492.828, Max-Change: 0.00445Iteration: 40, Log-Lik: -11492.808, Max-Change: 0.00444Iteration: 41, Log-Lik: -11492.780, Max-Change: 0.00432Iteration: 42, Log-Lik: -11492.754, Max-Change: 0.00574Iteration: 43, Log-Lik: -11492.725, Max-Change: 0.00402Iteration: 44, Log-Lik: -11492.704, Max-Change: 0.00383Iteration: 45, Log-Lik: -11492.685, Max-Change: 0.00367Iteration: 46, Log-Lik: -11492.597, Max-Change: 0.00571Iteration: 47, Log-Lik: -11492.588, Max-Change: 0.00215Iteration: 48, Log-Lik: -11492.581, Max-Change: 0.00250Iteration: 49, Log-Lik: -11492.553, Max-Change: 0.00391Iteration: 50, Log-Lik: -11492.548, Max-Change: 0.00151Iteration: 51, Log-Lik: -11492.545, Max-Change: 0.00167Iteration: 52, Log-Lik: -11492.541, Max-Change: 0.00149Iteration: 53, Log-Lik: -11492.538, Max-Change: 0.00179Iteration: 54, Log-Lik: -11492.535, Max-Change: 0.00136Iteration: 55, Log-Lik: -11492.529, Max-Change: 0.00117Iteration: 56, Log-Lik: -11492.528, Max-Change: 0.00086Iteration: 57, Log-Lik: -11492.526, Max-Change: 0.00085Iteration: 58, Log-Lik: -11492.521, Max-Change: 0.00062Iteration: 59, Log-Lik: -11492.520, Max-Change: 0.00059Iteration: 60, Log-Lik: -11492.520, Max-Change: 0.00055Iteration: 61, Log-Lik: -11492.517, Max-Change: 0.00014Iteration: 62, Log-Lik: -11492.517, Max-Change: 0.00013Iteration: 63, Log-Lik: -11492.517, Max-Change: 0.00014Iteration: 64, Log-Lik: -11492.516, Max-Change: 0.00089Iteration: 65, Log-Lik: -11492.516, Max-Change: 0.00011Iteration: 66, Log-Lik: -11492.516, Max-Change: 0.00054Iteration: 67, Log-Lik: -11492.516, Max-Change: 0.00070Iteration: 68, Log-Lik: -11492.515, Max-Change: 0.00020Iteration: 69, Log-Lik: -11492.515, Max-Change: 0.00009## F1 h2
## func_stand 0.895 0.801
## func_hous 0.924 0.854
## func_learn 0.826 0.682
## func_sociality 0.830 0.689
## func_emot 0.693 0.480
## func_con 0.708 0.501
## func_walk 0.883 0.780
## func_wash 0.961 0.924
## func_dress 0.963 0.928
## func_deal_strangers 0.774 0.598
## func_friend 0.786 0.618
## func_work 0.871 0.759
##
## SS loadings: 8.614
## Proportion Var: 0.718
##
## Factor correlations:
##
## F1
## F1 1## $items
## a b1 b2 b3 b4
## func_stand 3.412 0.151 0.832 1.386 2.072
## func_hous 4.117 0.804 1.446 1.930 2.520
## func_learn 2.495 1.216 1.929 2.482 3.229
## func_sociality 2.531 1.404 2.050 2.436 2.773
## func_emot 1.634 1.510 2.453 3.326 4.777
## func_con 1.705 1.229 2.343 3.285 5.289
## func_walk 3.205 0.022 0.702 1.296 1.902
## func_wash 5.951 1.374 1.813 2.202 2.662
## func_dress 6.090 1.388 1.823 2.315 2.879
## func_deal_strangers 2.077 2.203 2.941 3.548 4.133
## func_friend 2.166 2.256 2.887 3.488 3.905
## func_work 3.023 1.034 1.581 2.254 3.128
##
## $means
## F1
## 0
##
## $cov
## F1
## F1 1
## [1] "Correlation between WHODAS sum score and theta: 0.876"4. Distribution of WHODAS Items Responses
We now generate frequency tables, a distribution table (percentages), and a horizontal stacked bar chart for the WHODAS items.
4.1 Frequency and Distribution Table
## [1] "Frequency Tables for each WHODAS item:"## $func_stand
## .x
## 0 1 2 3 4 <NA>
## 1118 450 257 156 62 48
##
## $func_hous
## .x
## 0 1 2 3 4 <NA>
## 1570 290 108 53 20 50
##
## $func_learn
## .x
## 0 1 2 3 4 <NA>
## 1720 213 67 30 10 51
##
## $func_sociality
## .x
## 0 1 2 3 4 <NA>
## 1789 158 43 21 24 56
##
## $func_emot
## .x
## 0 1 2 3 4 <NA>
## 1734 206 69 26 3 53
##
## $func_con
## .x
## 0 1 2 3 4 <NA>
## 1653 286 73 25 1 53
##
## $func_walk
## .x
## 0 1 2 3 4 <NA>
## 1028 466 292 161 94 50
##
## $func_wash
## .x
## 0 1 2 3 4 <NA>
## 1859 107 42 22 11 50
##
## $func_dress
## .x
## 0 1 2 3 4 <NA>
## 1864 105 49 18 6 49
##
## $func_deal_strangers
## .x
## 0 1 2 3 4 <NA>
## 1948 66 18 6 3 50
##
## $func_friend
## .x
## 0 1 2 3 4 <NA>
## 1959 53 18 5 4 52
##
## $func_work
## .x
## 0 1 2 3 4 <NA>
## 1666 211 116 41 8 49## [1] "Distribution Table (Percentages) for WHODAS Items:"## # A tibble: 12 × 7
## item `No difficulty` `Mild difficulty` `Moderate difficulty`
## <chr> <dbl> <dbl> <dbl>
## 1 func_con 79.1 13.7 3.49
## 2 func_deal_strangers 93.2 3.16 0.861
## 3 func_dress 89.1 5.02 2.34
## 4 func_emot 82.9 9.85 3.30
## 5 func_friend 93.7 2.53 0.861
## 6 func_hous 75.1 13.9 5.16
## 7 func_learn 82.3 10.2 3.20
## 8 func_sociality 85.6 7.56 2.06
## 9 func_stand 53.5 21.5 12.3
## 10 func_walk 49.2 22.3 14.0
## 11 func_wash 88.9 5.12 2.01
## 12 func_work 79.7 10.1 5.55
## # ℹ 3 more variables: `Severe difficulty` <dbl>,
## # `Extreme difficulty or cannot do` <dbl>, `NA` <dbl>4.2 Stacked Bar Chart
5. WHODAS Comparison Items: Likert Plot
We now analyze the comparison items that capture respondents’ perceived change compared to before COVID. We recode the numeric responses (1 = Better, 2 = Same, 3 = Worse) into descriptive labels and create a Likert plot.
## The default 'loading.method' has changed to "revised" in {EGAnet} version >= 2.0.7.
##
## For the previous default (version <= 2.0.6), use `loading.method = "original"`## Loading Method: Revised
##
## 1
## func_dress 0.556
## func_friend 0.556
## func_wash 0.543
## func_learn 0.522
## func_hous 0.504
## func_sociality 0.452
## func_deal_strangers 0.449
## func_stand 0.448
## func_work 0.439
## func_walk 0.426
## func_con 0.352
## func_emot 0.286
## Standardized loadings >= |0.10| are displayed. To change this 'minimum', use `print(net.loads_object, minimum = 0.10)`## [1] "func_stand" "func_hous" "func_learn"
## [4] "func_sociality" "func_emot" "func_con"
## [7] "func_walk" "func_wash" "func_dress"
## [10] "func_deal_strangers" "func_friend" "func_work"## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^
## Warning: lavaan->ldw_parse_model_string():
## having identifiers with spaces ('Fat 1') is deprecated at line 1, pos 1
## Fat 1 =~ func_stand + func_hous + func_learn + func_sociality + func_emot + func_con + func_walk + func_wash + func_dress + func_deal_strangers + func_friend + func_work
## ^## Warning: lavaan->lav_data_full():
## some cases are empty and will be ignored: 14 19 39 43 58 60 84 112 114 144
## 171 174 179 186 202 217 356 439 464 468 485 548 603 914 1161 1341 1343
## 1345 1356 1396 1397 1404 1422 1471 1639 1658 1659 1720 1774 1814 2056 2065
## 2086 2091.## Registered S3 method overwritten by 'kutils':
## method from
## print.likert likert
## npar fmin
## 60.000 0.314
## chisq df
## 1283.540 54.000
## pvalue chisq.scaled
## 0.000 1383.566
## df.scaled pvalue.scaled
## 54.000 0.000
## chisq.scaling.factor baseline.chisq
## 0.943 184799.408
## baseline.df baseline.pvalue
## 66.000 0.000
## baseline.chisq.scaled baseline.df.scaled
## 69552.801 66.000
## baseline.pvalue.scaled baseline.chisq.scaling.factor
## 0.000 2.659
## cfi tli
## 0.993 0.992
## cfi.scaled tli.scaled
## 0.981 0.977
## cfi.robust tli.robust
## 0.648 0.570
## nnfi rfi
## 0.992 0.992
## nfi pnfi
## 0.993 0.812
## ifi rni
## 0.993 0.993
## nnfi.scaled rfi.scaled
## 0.977 0.976
## nfi.scaled pnfi.scaled
## 0.980 0.802
## ifi.scaled rni.scaled
## 0.981 0.981
## nnfi.robust rni.robust
## 0.570 0.648
## rmsea rmsea.ci.lower
## 0.105 0.101
## rmsea.ci.upper rmsea.ci.level
## 0.111 0.900
## rmsea.pvalue rmsea.close.h0
## 0.000 0.050
## rmsea.notclose.pvalue rmsea.notclose.h0
## 1.000 0.080
## rmsea.scaled rmsea.ci.lower.scaled
## 0.110 0.105
## rmsea.ci.upper.scaled rmsea.pvalue.scaled
## 0.115 0.000
## rmsea.notclose.pvalue.scaled rmsea.robust
## 1.000 0.320
## rmsea.ci.lower.robust rmsea.ci.upper.robust
## 0.295 0.344
## rmsea.pvalue.robust rmsea.notclose.pvalue.robust
## 0.000 1.000
## rmr rmr_nomean
## 0.081 0.103
## srmr srmr_bentler
## 0.103 0.081
## srmr_bentler_nomean crmr
## 0.103 0.085
## crmr_nomean srmr_mplus
## 0.112 NA
## srmr_mplus_nomean cn_05
## NA 116.014
## cn_01 gfi
## 130.226 0.994
## agfi pgfi
## 0.988 0.471
## mfi wrmr
## 0.740 3.355Interpretation
IRT Analysis:
The graded-response IRT model yielded item discrimination (a) and
threshold (b) parameters for each WHODAS item. Items with high
discrimination—such as func_wash (a = 5.95) and
func_dress (a = 6.09)—are particularly sensitive to differences
in the latent disability trait. Threshold parameters indicate the latent
trait levels at which respondents transition between categories (e.g.,
for func_stand, thresholds at 0.151, 0.832, 1.386, and 2.072
suggest progressively higher difficulty endorsements).
WHODAS Sum Score vs. Theta:
The latent trait (theta) derived from the IRT model shows a strong
positive correlation (r = 0.876) with the simple sum score of the WHODAS
items. This indicates that the sum score is a reasonable approximation
of the underlying disability construct, although theta offers a refined
measure that accounts for differential item characteristics.
Item Response Distributions:
Frequency and distribution tables, along with the stacked bar chart,
reveal variability across WHODAS items. For instance,
func_friend and func_deal_strangers display very high
percentages of “No difficulty” responses, whereas items like
func_stand and func_walk have a broader spread across
categories. This suggests that certain functional domains (e.g.,
mobility) are more affected in this sample.
Comparison Items (Likert Plot):
The Likert plot for the comparison items—where responses are recoded as
1 = Better, 2 = Same, 3 = Worse—indicates that most respondents perceive
little improvement, with many reporting that their current functioning
is “Same” or “Worse” compared to before COVID. This underscores the
generally negative impact of COVID-19 on patient functioning.