Variable Modifications

As discussed elsewhere, we have designed this website so that investigators can combine data across any or all of the 5 cold studies when the appropriate variables are available.  A major challenge in accomplishing this aim was that the same construct was sometimes represented differently in one or more of the studies.   For example, the anxious mood subscale that was computed from participants’ reports of daily emotional states while in quarantine was comprised of 4 items in PCS1, 3 items in PCS2, and 2 items in PMBC and PCS3.  It is important to remember that these studies were all designed individually, and not with the intent of being combined.  Measures for each study were chosen because they best represented our view of a variable at the time of designing a particular study or because they best fit the questions we were addressing in a particular study.  The idea of aggregating these studies did not arise until later. 

In order to use such variables in the aggregate data set, we needed to homogenize them, i.e., use the same definitions.  Consequently, there are key variables in the aggregate data set that are different than the same variables in the original individual data sets.  In most cases, the differences are trivial.  However, they are important for accurately defining variables used in aggregate analyses.  It is also important that investigators analyzing an individual data set (especially those replicating previously reported analyses) to use the original variables from the individual studies.

Below we describe some of the variable modifications that were made to establish equivalency across two or more studies in the Aggregated Data Set.  This information also is available in relevant sections of the Aggregated Data the codebook.

Psychological and Social Variables

State positive and negative affect were assessed in BCS, PCS2, and PMBC.   Because identical scales were not used, standardized (z-score) values were computed to create equivalency across the three studies. Variables representing the three positive and three negative subscales are available for PCS2 and PMBC; variables representing overall positive and negative affect are available for all 3 studies. All values are standardized.

Trait positive and negative affect and all component subscales were measured in PCS1, PCS2, and PCS3. For PCS1 and PCS2, the nearly all values included in the Aggregated data set represent the average scores across 2 (PCS1) or 3 (PCS2) administrations. The two exceptions are anger and overall negative affect. Items comprising the anger subscale were presented during only a single administration in PCS2, and thus overall negative affect was computed only for that administration. All values are in the original metric of the scale.

Marital quality.  The Clark Marital Intimacy Scale (CMI) was administered in PMBC and PCS3, with the former study using the full scale and the latter a short-form. To establish equivalency across the two studies, PMBC CMI scores were re-calculated using only the items included in the short-form version. All values are in the original metric of the scale.

Big 5 personality.  All of the Big Five personality factors were assessed in PCS1, PCS2, PMBC, and PCS3; only extraversion was measured in all 5 studies. PCS1, PCS2, and PMBC used Goldberg’s Adjective Scale1; PCS3 used the International Personality Item Pool (IPIP) Big 5 scale2, 3; and BCS used the Extraversion/Introversion items from the Eysenck Personality Questionnaire4. Because of the heterogeneity of the personality measures, standardized (z-score) values were computed to create equivalency across the studies. All values are standardized.

Optimism was measured in PCS1, PMBC, and PCS3. Whereas PCS1 used the original Life Orientation Test (LOT), the latter two studies used the revised LOT. To establish equivalency across the three studies, PCS1 LOT scores were re-calculated using only the items included in the revised version. All values are in the original metric of the scale.

Hostility.  The Cook Medley Hostility Scale was administered in PMBC and PCS3. Because the scale was given on 2 occasions in PMBC, the values included in the Aggregated data set represent the average of the 2 administrations. All values are in the original metric of the scale.

Social support.  The Interpersonal Support Evaluation List (ISEL) was administered in BCS, PCS2, PMBC, and PCS3. Whereas the BCS used the full 40-item scale, the three more recent studies used the 12-item version. To establish equivalency across the four studies, BCS ISEL scores were re-calculated using only the items that comprise the 12-item version.

Perceived stress.  The Perceived Stress Scale was administered in all five studies. Because all but one of the studies (BCS) administered the 10-item version of the scale, only the items comprising that version and the 10-item and 4-item totals are included in the Aggregated data set. All values are in the original metric of the scale.

Stressful life events.  Because of a printing error, response options for “sibling”, “child”, and “other relatives” appeared as a single response in the version of the Life Event List (LEL) that was administered in PCS3. To establish equivalency between the two studies, BCS “sibling”, “child”, and “other relative” response options were collapsed.

In PCS3, participants also had the options of identifying “spouse/partner’s parent”, “brother or sister”, and “friend” as persons affected by the three “additional life events”. Because these options were not available in BCS, they were collapsed with the “other” category for the Aggregated data set.

Daily Interviews

An important distinction between the daily interview data that were collected in PMBC and those collected in PCS3 is that PMBC participants were asked to report exclusively on activities in which they engaged with other persons. By comparison, PCS3 participants were asked to report all activities (social or solitary). To establish equivalency across the two studies, the activity variables presented here are comprised of social activities only.

Receipt and provision of social support were assessed in PCS2 and PMBC. In PCS2, participants were asked whether they gave or received support whereas in PMBC, participants were asked to rate the supportiveness of the other person if they shared something with that person and their own supportiveness if another person shared something with them. To create equivalent items, participants in PMBC were determined to have been engaged in a “support” interaction if the total support rating for the interaction was greater than or equal to 12 (possible range for supportiveness ratings = 0 to 16).

Saliva Cortisol

In order to establish equivalency of saliva cortisol data across PCS2, PMBC, and PCS3, the Aggregated Data Set includes data for only 7 of the 11 PCS2 pre-challenge saliva samples and for 8 of the 14 PCS2 Quarantine Day 0 samples.  Moreover, corresponding pre-challenge and day 0 area-under-the-curve (AUC) values appearing in the Aggregated data set were computed using 7 and 8 samples, respectively.  Thus, AUC values for PCS2 participants are not completely comparable across the stand-alone PCS2 and Aggregated data sets. 

Importantly, the PCS2 variables included in the aggregated data set were re-named to be consistent with variables from the other 2 studies.   For example, pre-challenge sample 10 (14 hours post-waking) in the stand-alone PCS2 data set has been renamed to sample 7 in the aggregate data set.   Table 1 below presents the original numbering scheme for the PCS2 pre-challenge samples, the time each sample was taken, the adjusted numbering scheme, and the analogous sampling time for saliva samples obtained in PMBC and PCS3.  Table 2 presents the same information for Quarantine Day 0.  In both tables, differences in timing of 30 minutes or more are indicated by bold type.

Table 1.  Changes in numbering of PCS2 pre-challenge cortisol saliva samples for incorporation into the aggregated data set.

Table 2.  Changes in numbering of PCS2 Quarantine Day 0 cortisol saliva samples for incorporation into the aggregated data set.

†In PMBC and PCS3, participants were instructed to wake by 8:00 am. However, there was some degree of variability.