Analyse van COVID-19 Symptomen bij Vrouwen met RRMS
Introductie
Onderzoeksgroep: Vrouwen met RRMS en COVID-19 Symptomen
In dit groepsproject hebben we onderzocht in hoeverre vrouwelijke patiënten met relapsing remitting multiple sclerose (RRMS) die bepaalde medicamenteuze therapieën ondergaan, meer gerapporteerde COVID-19 symptomen ervaren in vergelijking met andere behandelingen.
Hypotheses:
- H₀: Het gebruik van
specifieke medicatie heeft geen effect op de kans op het
ontwikkelen van bepaalde symptomen.
- Hₐ: Het
gebruik van specifieke medicatie heeft wél een effect op de kans
op het ontwikkelen van bepaalde symptomen.
Bijkomende onderzoeksvraag:
Zijn er bepaalde COVID-19 symptomen die vaker samen voorkomen?
Objectieven
- Het voorkomen van COVID-19 symptomen bij vrouwen met RRMS beschrijven.
- Het medicatiegebruik bij vrouwen met RRMS in kaart brengen.
- Het verband onderzoeken tussen medicatiegebruik en meer
gerapporteerde COVID-19 symptomen, gebruikmakend van de dataset
GDSI_OpenDataset_Final.
Groepsleden
- An-Sofie Verhoeven
- Dennis Hertoghe
- Dries Nassen
Methodologie
We hebben gewerkt in verschillende stappen om ons onderzoek systematisch uit te voeren:
1. Inladen van de data
De dataset GDSI_OpenDataset_Final werd geïmporteerd en
gecontroleerd op volledigheid.
2. Descriptieve analyse
Het aantal vrouwen met RRMS dat COVID-19 symptomen rapporteerde werd geanalyseerd, inclusief de verdeling van suspected en confirmed gevallen. Daarnaast werd medicatiegebruik in kaart gebracht.
3. Sorteren op enkel vrouwen met RRMS
De dataset werd gefilterd om alleen vrouwen met RRMS te behouden, aangezien deze groep het meest vertegenwoordigd is.
4. Imputatie van missende waarden
Missende antwoorden in de symptoomkolommen werden vervangen door “no” om uniformiteit in de dataset te krijgen.
5. Data voorbereiden voor clusteranalyse
Symptoomkolommen werden binair gemaakt (aanwezig = 1, afwezig = 0) om clustering mogelijk te maken.
6. Hiërarchische clustering
Symptomen werden geclusterd met een hiërarchische methode, resulterend in een dendrogram om de samenhang tussen symptomen te visualiseren.
7. Statistische toetsing
Een Chi-kwadraat test voor goodness of fit werd uitgevoerd om te beoordelen of de clustering van symptomen significant is.
8. Clusterinformatie toevoegen
Een extra kolom werd toegevoegd aan de dataset met de clusterinformatie per patiënt.
9. Logistische regressie
Per symptoom werd een logistische regressie uitgevoerd om te onderzoeken welke factoren de kans op het optreden van een symptoom beïnvloeden.
10. Discussie en interpretatie
De resultaten van de analyses werden geïnterpreteerd en besproken.
Descriptieve Analyse
We hebben de data van vrouwen met RRMS geanalyseerd om inzicht te krijgen in hun COVID-19 symptomen en medicatiegebruik. Hieronder worden de resultaten weergegeven.
Tabellen
| Symptoomstatus | Aantal |
|---|---|
| 7 | |
| no | 489 |
| yes | 243 |
| Type Medicatie | Aantal |
|---|---|
| 82 | |
| currently not using any DMT | 59 |
| currently on alemtuzumab | 8 |
| currently on another drug not listed | 117 |
| currently on cladribine | 29 |
| currently on dimethyl fumarate | 101 |
| currently on fingolimod | 90 |
| currently on glatiramer | 43 |
| currently on interferon | 59 |
| currently on natalizumab | 49 |
| currently on ocrelizumab | 52 |
| currently on rituximab | 8 |
| currently on teriflunomide | 42 |
| Symptoom | Aantal ‘Yes’ |
|---|---|
| Droge hoest | 139 |
| Rillingen | 74 |
| Vermoeidheid | 159 |
| Kortademigheid | 70 |
| Keelpijn | 114 |
| Koorts | 95 |
| Verlies van reuk/smaak | 60 |
| Pijn | 125 |
| Longontsteking | 10 |
| Neusverstopping | 115 |
Visualisatie van afstandmatrix
# transpose
symptomen_transposed <- data.table::transpose(as.data.table(vrouwenRRMS_symptomen),keep.names = "col")
# gebruik eerste kolom (omschrijving van het symptoom) als rownames
symptomen_transposed <- data.frame(symptomen_transposed, row.names = 1)
rm(vrouwenRRMS_symptomen) # tabel niet meer nodig
#afstand berekenen WAAROM binary!!!!
distance_matrix <- dist(symptomen_transposed, method = "binary")
fviz_dist(distance_matrix)
Hierargisch clusteren
#cluster methode
hclust_result <- hclust(distance_matrix, method = "ward.D")
plot(hclust_result, main = "Dendrogram van Hiërarchische Clustering voor COVID-19 Symptomen",labels = rownames(symptomen_transposed),
cex=0.7)
Cluster frequenties
| Cluster | Frequency | Proportion |
|---|---|---|
| 1 | 3 | 0.3 |
| 2 | 5 | 0.5 |
| 3 | 1 | 0.1 |
| 4 | 1 | 0.1 |
Histogram cluster frequenties
Statistisch testen van clusters
## Warning in stats::chisq.test(x, y, ...): Chi-squared approximation may be
## incorrect| Statistiek | p.waarde | Vrijheidsgraden | |
|---|---|---|---|
| X-squared | 4.4 | 0.2213854 | 3 |
Logistische regressie
Symptoom rillingen
modelchills <- glm(chills ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = chills ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -37.7215 2433.4455
## dmt_type_overallcurrently not using any DMT 0.7388 0.8001
## dmt_type_overallcurrently on alemtuzumab 15.1555 17212.7910
## dmt_type_overallcurrently on another drug not listed 0.1201 0.7571
## dmt_type_overallcurrently on cladribine 0.6383 1.0806
## dmt_type_overallcurrently on dimethyl fumarate -0.3634 0.7861
## dmt_type_overallcurrently on fingolimod -0.4860 0.7933
## dmt_type_overallcurrently on glatiramer 0.4678 0.9101
## dmt_type_overallcurrently on interferon 0.6457 0.9762
## dmt_type_overallcurrently on natalizumab -0.1613 0.9179
## dmt_type_overallcurrently on ocrelizumab -1.4340 0.9122
## dmt_type_overallcurrently on rituximab 0.2442 1.5575
## dmt_type_overallcurrently on teriflunomide 0.1229 0.9582
## symptoom_cluster1 20.3459 1681.9380
## symptoom_cluster2 17.1315 1758.6197
## symptoom_cluster3 1.0268 0.4480
## z value Pr(>|z|)
## (Intercept) -0.016 0.9876
## dmt_type_overallcurrently not using any DMT 0.923 0.3558
## dmt_type_overallcurrently on alemtuzumab 0.001 0.9993
## dmt_type_overallcurrently on another drug not listed 0.159 0.8739
## dmt_type_overallcurrently on cladribine 0.591 0.5548
## dmt_type_overallcurrently on dimethyl fumarate -0.462 0.6439
## dmt_type_overallcurrently on fingolimod -0.613 0.5401
## dmt_type_overallcurrently on glatiramer 0.514 0.6073
## dmt_type_overallcurrently on interferon 0.661 0.5083
## dmt_type_overallcurrently on natalizumab -0.176 0.8605
## dmt_type_overallcurrently on ocrelizumab -1.572 0.1159
## dmt_type_overallcurrently on rituximab 0.157 0.8754
## dmt_type_overallcurrently on teriflunomide 0.128 0.8980
## symptoom_cluster1 0.012 0.9903
## symptoom_cluster2 0.010 0.9922
## symptoom_cluster3 2.292 0.0219 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 480.91 on 738 degrees of freedom
## Residual deviance: 188.15 on 723 degrees of freedom
## AIC: 220.15
##
## Number of Fisher Scoring iterations: 21Symptoom droge hoest
modeldrycough <- glm(dry_cough ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = dry_cough ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -31.2872 1514.4174
## dmt_type_overallcurrently not using any DMT -0.3257 0.7053
## dmt_type_overallcurrently on alemtuzumab 9.7212 10445.5918
## dmt_type_overallcurrently on another drug not listed -0.4952 0.6388
## dmt_type_overallcurrently on cladribine -0.6878 0.9113
## dmt_type_overallcurrently on dimethyl fumarate -1.1560 0.6827
## dmt_type_overallcurrently on fingolimod -0.6944 0.6458
## dmt_type_overallcurrently on glatiramer 0.7852 0.9360
## dmt_type_overallcurrently on interferon -1.5712 0.7414
## dmt_type_overallcurrently on natalizumab -1.1202 0.6984
## dmt_type_overallcurrently on ocrelizumab 0.2889 0.7740
## dmt_type_overallcurrently on rituximab 14.5448 1071.8416
## dmt_type_overallcurrently on teriflunomide -1.0404 0.7895
## symptoom_cluster1 0.4997 0.3069
## symptoom_cluster2 31.8901 1514.4174
## symptoom_cluster3 0.6295 0.3636
## z value Pr(>|z|)
## (Intercept) -0.021 0.9835
## dmt_type_overallcurrently not using any DMT -0.462 0.6443
## dmt_type_overallcurrently on alemtuzumab 0.001 0.9993
## dmt_type_overallcurrently on another drug not listed -0.775 0.4382
## dmt_type_overallcurrently on cladribine -0.755 0.4504
## dmt_type_overallcurrently on dimethyl fumarate -1.693 0.0904 .
## dmt_type_overallcurrently on fingolimod -1.075 0.2823
## dmt_type_overallcurrently on glatiramer 0.839 0.4015
## dmt_type_overallcurrently on interferon -2.119 0.0341 *
## dmt_type_overallcurrently on natalizumab -1.604 0.1087
## dmt_type_overallcurrently on ocrelizumab 0.373 0.7090
## dmt_type_overallcurrently on rituximab 0.014 0.9892
## dmt_type_overallcurrently on teriflunomide -1.318 0.1875
## symptoom_cluster1 1.628 0.1035
## symptoom_cluster2 0.021 0.9832
## symptoom_cluster3 1.731 0.0834 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 714.53 on 738 degrees of freedom
## Residual deviance: 272.49 on 723 degrees of freedom
## AIC: 304.49
##
## Number of Fisher Scoring iterations: 20Symptoom vermoeidheid
modelfatigue <- glm(fatigue ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = fatigue ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -3.078e+01 1.510e+03
## dmt_type_overallcurrently not using any DMT -7.616e-01 7.585e-01
## dmt_type_overallcurrently on alemtuzumab 9.217e+00 1.044e+04
## dmt_type_overallcurrently on another drug not listed 2.738e-01 7.443e-01
## dmt_type_overallcurrently on cladribine -5.066e-01 1.025e+00
## dmt_type_overallcurrently on dimethyl fumarate -1.031e+00 7.485e-01
## dmt_type_overallcurrently on fingolimod -9.853e-02 7.320e-01
## dmt_type_overallcurrently on glatiramer -8.108e-01 8.272e-01
## dmt_type_overallcurrently on interferon 1.532e-01 8.745e-01
## dmt_type_overallcurrently on natalizumab 2.446e-01 8.163e-01
## dmt_type_overallcurrently on ocrelizumab -2.242e+00 7.747e-01
## dmt_type_overallcurrently on rituximab 1.405e+01 1.069e+03
## dmt_type_overallcurrently on teriflunomide -1.421e+00 8.377e-01
## symptoom_cluster1 9.504e-01 3.386e-01
## symptoom_cluster2 3.143e+01 1.510e+03
## symptoom_cluster3 8.376e-01 4.132e-01
## z value Pr(>|z|)
## (Intercept) -0.020 0.98374
## dmt_type_overallcurrently not using any DMT -1.004 0.31532
## dmt_type_overallcurrently on alemtuzumab 0.001 0.99930
## dmt_type_overallcurrently on another drug not listed 0.368 0.71303
## dmt_type_overallcurrently on cladribine -0.494 0.62116
## dmt_type_overallcurrently on dimethyl fumarate -1.378 0.16830
## dmt_type_overallcurrently on fingolimod -0.135 0.89292
## dmt_type_overallcurrently on glatiramer -0.980 0.32699
## dmt_type_overallcurrently on interferon 0.175 0.86095
## dmt_type_overallcurrently on natalizumab 0.300 0.76444
## dmt_type_overallcurrently on ocrelizumab -2.894 0.00381 **
## dmt_type_overallcurrently on rituximab 0.013 0.98951
## dmt_type_overallcurrently on teriflunomide -1.697 0.08976 .
## symptoom_cluster1 2.807 0.00500 **
## symptoom_cluster2 0.021 0.98340
## symptoom_cluster3 2.027 0.04265 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 769.61 on 738 degrees of freedom
## Residual deviance: 235.20 on 723 degrees of freedom
## AIC: 267.2
##
## Number of Fisher Scoring iterations: 20Symptoom koorts
modelfever <- glm(fever ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)
summary(modelfever)##
## Call:
## glm(formula = fever ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -2.214e+01 1.191e+03
## dmt_type_overallcurrently not using any DMT 2.734e-01 7.830e-01
## dmt_type_overallcurrently on alemtuzumab 5.695e-01 1.040e+04
## dmt_type_overallcurrently on another drug not listed 1.122e+00 7.983e-01
## dmt_type_overallcurrently on cladribine -8.273e-02 1.043e+00
## dmt_type_overallcurrently on dimethyl fumarate 3.209e-01 7.939e-01
## dmt_type_overallcurrently on fingolimod 1.028e+00 8.134e-01
## dmt_type_overallcurrently on glatiramer 6.096e-02 9.028e-01
## dmt_type_overallcurrently on interferon 6.628e-01 9.796e-01
## dmt_type_overallcurrently on natalizumab -9.474e-04 9.026e-01
## dmt_type_overallcurrently on ocrelizumab 1.435e+00 9.999e-01
## dmt_type_overallcurrently on rituximab 2.643e-01 1.560e+00
## dmt_type_overallcurrently on teriflunomide 1.392e-01 9.616e-01
## symptoom_cluster1 2.280e+01 1.191e+03
## symptoom_cluster2 -9.257e-01 1.247e+00
## symptoom_cluster3 1.063e+00 4.698e-01
## z value Pr(>|z|)
## (Intercept) -0.019 0.9852
## dmt_type_overallcurrently not using any DMT 0.349 0.7270
## dmt_type_overallcurrently on alemtuzumab 0.000 1.0000
## dmt_type_overallcurrently on another drug not listed 1.405 0.1599
## dmt_type_overallcurrently on cladribine -0.079 0.9368
## dmt_type_overallcurrently on dimethyl fumarate 0.404 0.6860
## dmt_type_overallcurrently on fingolimod 1.263 0.2064
## dmt_type_overallcurrently on glatiramer 0.068 0.9462
## dmt_type_overallcurrently on interferon 0.677 0.4987
## dmt_type_overallcurrently on natalizumab -0.001 0.9992
## dmt_type_overallcurrently on ocrelizumab 1.436 0.1511
## dmt_type_overallcurrently on rituximab 0.169 0.8654
## dmt_type_overallcurrently on teriflunomide 0.145 0.8849
## symptoom_cluster1 0.019 0.9847
## symptoom_cluster2 -0.743 0.4578
## symptoom_cluster3 2.264 0.0236 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 567.00 on 738 degrees of freedom
## Residual deviance: 185.45 on 723 degrees of freedom
## AIC: 217.45
##
## Number of Fisher Scoring iterations: 20Symptoom reukverlies
modellossofsmell <- glm(loss_smell_taste ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: algorithm did not converge##
## Call:
## glm(formula = loss_smell_taste ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -2.657e+01 4.000e+04
## dmt_type_overallcurrently not using any DMT -3.548e-13 6.110e+04
## dmt_type_overallcurrently on alemtuzumab -9.917e-13 1.321e+05
## dmt_type_overallcurrently on another drug not listed -7.615e-13 5.140e+04
## dmt_type_overallcurrently on cladribine -2.463e-12 7.699e+04
## dmt_type_overallcurrently on dimethyl fumarate -7.915e-13 5.303e+04
## dmt_type_overallcurrently on fingolimod -1.908e-12 5.457e+04
## dmt_type_overallcurrently on glatiramer -6.543e-13 6.723e+04
## dmt_type_overallcurrently on interferon -2.619e-12 6.082e+04
## dmt_type_overallcurrently on natalizumab -1.856e-12 6.470e+04
## dmt_type_overallcurrently on ocrelizumab 3.301e-13 6.327e+04
## dmt_type_overallcurrently on rituximab -1.729e-12 1.321e+05
## dmt_type_overallcurrently on teriflunomide -1.419e-12 6.767e+04
## symptoom_cluster1 8.463e-12 4.872e+04
## symptoom_cluster2 -5.592e-12 4.425e+04
## symptoom_cluster3 5.313e+01 5.358e+04
## z value Pr(>|z|)
## (Intercept) -0.001 0.999
## dmt_type_overallcurrently not using any DMT 0.000 1.000
## dmt_type_overallcurrently on alemtuzumab 0.000 1.000
## dmt_type_overallcurrently on another drug not listed 0.000 1.000
## dmt_type_overallcurrently on cladribine 0.000 1.000
## dmt_type_overallcurrently on dimethyl fumarate 0.000 1.000
## dmt_type_overallcurrently on fingolimod 0.000 1.000
## dmt_type_overallcurrently on glatiramer 0.000 1.000
## dmt_type_overallcurrently on interferon 0.000 1.000
## dmt_type_overallcurrently on natalizumab 0.000 1.000
## dmt_type_overallcurrently on ocrelizumab 0.000 1.000
## dmt_type_overallcurrently on rituximab 0.000 1.000
## dmt_type_overallcurrently on teriflunomide 0.000 1.000
## symptoom_cluster1 0.000 1.000
## symptoom_cluster2 0.000 1.000
## symptoom_cluster3 0.001 0.999
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4.1631e+02 on 738 degrees of freedom
## Residual deviance: 4.2874e-09 on 723 degrees of freedom
## AIC: 32
##
## Number of Fisher Scoring iterations: 25Column
Symptoom verstopte neus
modelnasalcongestion <- glm(nasal_congestion ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)
summary(modelnasalcongestion)##
## Call:
## glm(formula = nasal_congestion ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -2.034e+01 7.782e+02
## dmt_type_overallcurrently not using any DMT -3.410e-01 6.251e-01
## dmt_type_overallcurrently on alemtuzumab -2.231e-01 6.317e+03
## dmt_type_overallcurrently on another drug not listed 1.386e-01 5.810e-01
## dmt_type_overallcurrently on cladribine -9.410e-01 8.636e-01
## dmt_type_overallcurrently on dimethyl fumarate 1.955e-01 6.348e-01
## dmt_type_overallcurrently on fingolimod -9.843e-01 5.969e-01
## dmt_type_overallcurrently on glatiramer -6.833e-01 7.139e-01
## dmt_type_overallcurrently on interferon -6.311e-01 6.817e-01
## dmt_type_overallcurrently on natalizumab -7.596e-01 6.535e-01
## dmt_type_overallcurrently on ocrelizumab -2.040e-04 6.466e-01
## dmt_type_overallcurrently on rituximab -4.257e-01 1.497e+00
## dmt_type_overallcurrently on teriflunomide -3.898e-01 7.410e-01
## symptoom_cluster1 1.687e-01 2.952e-01
## symptoom_cluster2 2.060e+01 7.782e+02
## symptoom_cluster3 1.218e-01 3.282e-01
## z value Pr(>|z|)
## (Intercept) -0.026 0.9791
## dmt_type_overallcurrently not using any DMT -0.545 0.5854
## dmt_type_overallcurrently on alemtuzumab 0.000 1.0000
## dmt_type_overallcurrently on another drug not listed 0.239 0.8115
## dmt_type_overallcurrently on cladribine -1.090 0.2759
## dmt_type_overallcurrently on dimethyl fumarate 0.308 0.7581
## dmt_type_overallcurrently on fingolimod -1.649 0.0991 .
## dmt_type_overallcurrently on glatiramer -0.957 0.3385
## dmt_type_overallcurrently on interferon -0.926 0.3545
## dmt_type_overallcurrently on natalizumab -1.162 0.2451
## dmt_type_overallcurrently on ocrelizumab 0.000 0.9997
## dmt_type_overallcurrently on rituximab -0.284 0.7762
## dmt_type_overallcurrently on teriflunomide -0.526 0.5989
## symptoom_cluster1 0.572 0.5676
## symptoom_cluster2 0.026 0.9789
## symptoom_cluster3 0.371 0.7105
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 638.98 on 738 degrees of freedom
## Residual deviance: 299.38 on 723 degrees of freedom
## AIC: 331.38
##
## Number of Fisher Scoring iterations: 19Symptoom pijn
modelpain <- glm(pain ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = pain ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 +
## symptoom_cluster3, family = "binomial", data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -32.5505 1526.0814
## dmt_type_overallcurrently not using any DMT 0.7734 0.6429
## dmt_type_overallcurrently on alemtuzumab 10.9844 10447.2893
## dmt_type_overallcurrently on another drug not listed 0.5584 0.5811
## dmt_type_overallcurrently on cladribine 0.1313 0.8498
## dmt_type_overallcurrently on dimethyl fumarate 0.9632 0.6490
## dmt_type_overallcurrently on fingolimod 0.6053 0.5941
## dmt_type_overallcurrently on glatiramer 0.2403 0.7163
## dmt_type_overallcurrently on interferon -0.2193 0.6951
## dmt_type_overallcurrently on natalizumab 1.2438 0.6825
## dmt_type_overallcurrently on ocrelizumab -0.6662 0.6749
## dmt_type_overallcurrently on rituximab 15.7930 1080.2646
## dmt_type_overallcurrently on teriflunomide 0.9660 0.7747
## symptoom_cluster1 0.5797 0.3009
## symptoom_cluster2 31.8408 1526.0813
## symptoom_cluster3 0.3805 0.3403
## z value Pr(>|z|)
## (Intercept) -0.021 0.9830
## dmt_type_overallcurrently not using any DMT 1.203 0.2290
## dmt_type_overallcurrently on alemtuzumab 0.001 0.9992
## dmt_type_overallcurrently on another drug not listed 0.961 0.3366
## dmt_type_overallcurrently on cladribine 0.155 0.8772
## dmt_type_overallcurrently on dimethyl fumarate 1.484 0.1378
## dmt_type_overallcurrently on fingolimod 1.019 0.3083
## dmt_type_overallcurrently on glatiramer 0.335 0.7373
## dmt_type_overallcurrently on interferon -0.316 0.7523
## dmt_type_overallcurrently on natalizumab 1.822 0.0684 .
## dmt_type_overallcurrently on ocrelizumab -0.987 0.3235
## dmt_type_overallcurrently on rituximab 0.015 0.9883
## dmt_type_overallcurrently on teriflunomide 1.247 0.2124
## symptoom_cluster1 1.927 0.0540 .
## symptoom_cluster2 0.021 0.9834
## symptoom_cluster3 1.118 0.2635
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 671.80 on 738 degrees of freedom
## Residual deviance: 286.04 on 723 degrees of freedom
## AIC: 318.04
##
## Number of Fisher Scoring iterations: 20Symptoom pneumonie
modelpneumonia <- glm(pneumonia ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = pneumonia ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -36.3295 3840.4430
## dmt_type_overallcurrently not using any DMT -19.6104 8882.2680
## dmt_type_overallcurrently on alemtuzumab 12.7635 28355.3396
## dmt_type_overallcurrently on another drug not listed 0.6760 1.2577
## dmt_type_overallcurrently on cladribine -19.1139 12278.3740
## dmt_type_overallcurrently on dimethyl fumarate -0.3252 1.4721
## dmt_type_overallcurrently on fingolimod 0.8992 1.2963
## dmt_type_overallcurrently on glatiramer 0.3834 1.5231
## dmt_type_overallcurrently on interferon -18.5547 8482.6306
## dmt_type_overallcurrently on natalizumab -18.9422 9538.3917
## dmt_type_overallcurrently on ocrelizumab -19.3156 9235.5279
## dmt_type_overallcurrently on rituximab -19.1561 23736.6538
## dmt_type_overallcurrently on teriflunomide 0.6742 1.5523
## symptoom_cluster1 18.5375 2677.3437
## symptoom_cluster2 15.1054 2753.3333
## symptoom_cluster3 0.4590 0.8074
## z value Pr(>|z|)
## (Intercept) -0.009 0.992
## dmt_type_overallcurrently not using any DMT -0.002 0.998
## dmt_type_overallcurrently on alemtuzumab 0.000 1.000
## dmt_type_overallcurrently on another drug not listed 0.537 0.591
## dmt_type_overallcurrently on cladribine -0.002 0.999
## dmt_type_overallcurrently on dimethyl fumarate -0.221 0.825
## dmt_type_overallcurrently on fingolimod 0.694 0.488
## dmt_type_overallcurrently on glatiramer 0.252 0.801
## dmt_type_overallcurrently on interferon -0.002 0.998
## dmt_type_overallcurrently on natalizumab -0.002 0.998
## dmt_type_overallcurrently on ocrelizumab -0.002 0.998
## dmt_type_overallcurrently on rituximab -0.001 0.999
## dmt_type_overallcurrently on teriflunomide 0.434 0.664
## symptoom_cluster1 0.007 0.994
## symptoom_cluster2 0.005 0.996
## symptoom_cluster3 0.569 0.570
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 105.918 on 738 degrees of freedom
## Residual deviance: 61.428 on 723 degrees of freedom
## AIC: 93.428
##
## Number of Fisher Scoring iterations: 22Symptoom keelpijn
modelsorethroat <- glm(sore_throat ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = sore_throat ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -3.207e+01 1.520e+03
## dmt_type_overallcurrently not using any DMT 2.374e-02 6.254e-01
## dmt_type_overallcurrently on alemtuzumab 1.050e+01 1.045e+04
## dmt_type_overallcurrently on another drug not listed -3.580e-01 5.722e-01
## dmt_type_overallcurrently on cladribine -1.308e+00 9.381e-01
## dmt_type_overallcurrently on dimethyl fumarate -4.385e-01 6.212e-01
## dmt_type_overallcurrently on fingolimod -1.105e-01 5.858e-01
## dmt_type_overallcurrently on glatiramer -1.693e-02 7.129e-01
## dmt_type_overallcurrently on interferon -1.049e+00 7.077e-01
## dmt_type_overallcurrently on natalizumab -1.477e-01 6.473e-01
## dmt_type_overallcurrently on ocrelizumab -6.149e-01 6.453e-01
## dmt_type_overallcurrently on rituximab 1.532e+01 1.076e+03
## dmt_type_overallcurrently on teriflunomide 1.318e+00 9.016e-01
## symptoom_cluster1 -1.102e-01 2.981e-01
## symptoom_cluster2 3.252e+01 1.520e+03
## symptoom_cluster3 -4.485e-01 3.316e-01
## z value Pr(>|z|)
## (Intercept) -0.021 0.983
## dmt_type_overallcurrently not using any DMT 0.038 0.970
## dmt_type_overallcurrently on alemtuzumab 0.001 0.999
## dmt_type_overallcurrently on another drug not listed -0.626 0.532
## dmt_type_overallcurrently on cladribine -1.395 0.163
## dmt_type_overallcurrently on dimethyl fumarate -0.706 0.480
## dmt_type_overallcurrently on fingolimod -0.189 0.850
## dmt_type_overallcurrently on glatiramer -0.024 0.981
## dmt_type_overallcurrently on interferon -1.483 0.138
## dmt_type_overallcurrently on natalizumab -0.228 0.819
## dmt_type_overallcurrently on ocrelizumab -0.953 0.341
## dmt_type_overallcurrently on rituximab 0.014 0.989
## dmt_type_overallcurrently on teriflunomide 1.462 0.144
## symptoom_cluster1 -0.370 0.712
## symptoom_cluster2 0.021 0.983
## symptoom_cluster3 -1.352 0.176
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 635.59 on 738 degrees of freedom
## Residual deviance: 292.88 on 723 degrees of freedom
## AIC: 324.88
##
## Number of Fisher Scoring iterations: 20Symptoom kortademigheid
modelshortnessofbreath <- glm(shortness_breath ~ dmt_type_overall + symptoom_cluster1 + symptoom_cluster2 + symptoom_cluster3, family = "binomial", data = vrouwenRRMS)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred##
## Call:
## glm(formula = shortness_breath ~ dmt_type_overall + symptoom_cluster1 +
## symptoom_cluster2 + symptoom_cluster3, family = "binomial",
## data = vrouwenRRMS)
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -33.2050 2333.0830
## dmt_type_overallcurrently not using any DMT -0.4768 0.7810
## dmt_type_overallcurrently on alemtuzumab 10.6389 17198.8893
## dmt_type_overallcurrently on another drug not listed -0.9213 0.7627
## dmt_type_overallcurrently on cladribine -0.0928 1.0698
## dmt_type_overallcurrently on dimethyl fumarate -1.2596 0.7919
## dmt_type_overallcurrently on fingolimod -1.4977 0.8045
## dmt_type_overallcurrently on glatiramer -0.4326 0.9088
## dmt_type_overallcurrently on interferon -1.3804 0.9785
## dmt_type_overallcurrently on natalizumab -0.6553 0.9088
## dmt_type_overallcurrently on ocrelizumab -1.8365 0.9108
## dmt_type_overallcurrently on rituximab 15.5987 1651.5823
## dmt_type_overallcurrently on teriflunomide -1.9698 1.0451
## symptoom_cluster1 33.2772 2333.0832
## symptoom_cluster2 0.8412 1.2358
## symptoom_cluster3 -0.3680 0.4219
## z value Pr(>|z|)
## (Intercept) -0.014 0.9886
## dmt_type_overallcurrently not using any DMT -0.610 0.5416
## dmt_type_overallcurrently on alemtuzumab 0.001 0.9995
## dmt_type_overallcurrently on another drug not listed -1.208 0.2271
## dmt_type_overallcurrently on cladribine -0.087 0.9309
## dmt_type_overallcurrently on dimethyl fumarate -1.591 0.1117
## dmt_type_overallcurrently on fingolimod -1.862 0.0626 .
## dmt_type_overallcurrently on glatiramer -0.476 0.6341
## dmt_type_overallcurrently on interferon -1.411 0.1583
## dmt_type_overallcurrently on natalizumab -0.721 0.4709
## dmt_type_overallcurrently on ocrelizumab -2.016 0.0438 *
## dmt_type_overallcurrently on rituximab 0.009 0.9925
## dmt_type_overallcurrently on teriflunomide -1.885 0.0595 .
## symptoom_cluster1 0.014 0.9886
## symptoom_cluster2 0.681 0.4961
## symptoom_cluster3 -0.872 0.3830
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 463.10 on 738 degrees of freedom
## Residual deviance: 192.88 on 723 degrees of freedom
## AIC: 224.88
##
## Number of Fisher Scoring iterations: 21Discussie resultaat
Onze descriptieve analyses zorgden ervoor dat we de dataset konden vereenvoudigen naar enkel vrouwen met relapsing remitting multiple sclerose (RRMS). Hieruit kwam de vraag of dat vrouwen met RRMS bepaalde symptomen meer vertonen indien ze een specifieke medicamenteuze behandeling ondergaan. Deze symptomen varieerden van vermoeidheid en keelpijn tot ernstige symptomen zoals kortademigheid en pneumonie.
De clustering liet zien dat er een groepering van symptomen was met een resultaat van 4 clusters. Deze clusteranalyse was niet significant ten opzichte van de voorspelde waardes. Dit wil zeggen dat dat clusters die we gevormd hebben niet verschillen van een willekeurige verdeling. Redenen hiervoor kunnen zijn:
Binaire data zijn minder geschikt voor de Ward’s methode omdat deze beter met euclidische afstanden berekend wordt. De manier waarop Ward afstanden tussen gegevens berekent, komt dus niet goed overeen met binaire data, die moeilijk om te vormen zijn naar een continue en meetbare afstanden. Voor deze specifieke data zouden complete en average linkage methoden geschikter zijn geweest.
We hebben enkel gekeken of er covid-19 symptomen aanwezig waren. Alle patienten waren dus “suspected” of “confirmed” cases van covid. Hieruit volgt dat we geen conclusies kunnen trekken uit onze data waarbij we de kans op COVID kunnen voorspellen. Variabele (“not answered”) werden omgezet naar (“no”) wat mogelijks heeft geleidt tot onderschatting van symptomen. Daarnaast is er een observationele bias aangezien de studie ook zelf gerapporteerde symptomen bevatten waarbij subjectieve interpretaties kunnen ontstaan.
